A User's Guide

Oskar Pearson

Table of Contents
1. History and Credits
History and Credits
2. Installing Squid
Hardware Requirements
Gathering statistics
Hard Disks
RAM requirements
CPU Power
Choosing an Operating System
Basic System Setup
Default Squid directory structure
User and Group IDs
Getting Squid
Getting the Squid source code
Getting Binary Versions of Squid
Compiling Squid
Compilation Tools
Unpacking the Source Archive
Compilation options
Running configure
Compiling the Squid Source
Installing the Squid binary
3. Squid Configuration Basics
Version Control Systems
The Configuration File
Setting Squid's HTTP Port
Using Port 80
Email for the Cache Administrator
Effective User and Group ID
FTP login information
Access Control Lists and Access Control Operators
Simple Access Control
Ensuring Direct Access to Internal Machines
Communicating with other proxy servers
Your ISP's cache
Firewall Interactions
4. Starting Squid
Before Running Squid
Subdirectory Permissions
Running Squid
Testing Squid
Testing a Cache or Proxy Server with Client
Addition to Startup Files
5. Browser Configuration
Basic Configuration
Advanced Configuration
Basic Configuration
Host name
Browser-cache Interaction
Testing the Cache
Cache Auto-config
Web server config changes for autoconfig files
Autoconfig Script Coding
Super Proxy Script
cgi generated autoconfig files
Future directions
Ready to Go
6. Access Control and Access Control Operators
Uses of ACLs
Access Classes and Operators
Acl lines
A unique name
Decision String
Types of acl
Acl-operator lines
The other Acl-operators
SNMP Configuration
Querying the Squid SNMP server on port 3401
Running multiple SNMP servers on a cache machine
Delay Classes
Slowing down access to specific URLs
The Second Pool Class
The Second Pool Class
The Third Pool Class
Using Delay Pools in Real Life
7. Cache Hierarchies
Why Peer
Peer Configuration
The cache_peer Option
Peer Selection
Selecting by Destination Domain
Selecting with Acls
Other Peering Options
Multicast Cache Communication
Getting your machine ready for Multicast
Querying a Multicast Cache
Accepting Multicast Queries: The mcast_groups option
Other Multicast Cache Options
Cache Digests
Cache Hierarchy Structures
Two Peering Caches
Load Balancing Servers
The Cache Array Routing Protocol (CARP)
8. Accelerator Mode
When to use Accelerator Mode
Acceleration of a slow server
Replacing a combination cache/web server with Squid
Transparent Caching
Accelerator Configuration Options
The httpd_accel_host option
The httpd_accel_port option
The httpd_accel_with_proxy option
The httpd_accel_uses_host_header option
Related Configuration Options
The redirect_rewrites_host_header option
Refresh patterns
Access Control
Example Configurations
Replacing a Combination Web/Cache server
Accelerating Requests to a Slow Server
9. Transparent Caching
The Problem with Transparency
The Transparent Caching Process
Some Routing Basics
Packet Flow with Transparent Caches
Network Layout
Filtering Traffic
Unix machines
Routers (not done)
Layer-Four Switches (not done)
Kernel Redirection (not done)
Squid Settings (not done)
10. Not Yet Done: Squid Config files and options
11. Overall Layout (for writers)
12. GNU Free Documentation License
GNU Free Documentation License
List of Examples
3-1. Effective User and Group IDs
3-2. Theoretical Access List
3-3. Access Lists using Classes
3-4. CIDR vs Netmask Source-IP Notation
3-5. Example Complete ACL list
3-6. Using always and never_direct
4-1. Using the -h and -p client Options
4-2. Retrieving Pages directly from a remote site with client
4-3. Printing timing information for a page download
4-4. Accessing a site through the cache
4-5. Runcache command in the startup files
5-1. Restarting Apache
5-2. A very basic autoconfig file
5-3. Connecting to a cache server
5-4. Connecting to a cache server, with failover
5-5. dnsDomainIs
5-6. Using multiple dnsDomainIs calls
5-7. using the isInNet call
5-8. using isPlainHostName to decide if the connection should be direct
5-9. myIpAddress
5-10. shExpMatch
5-11. url.substring
5-12. A small organization's proxy config file
5-13. Dialup ISP autoconfig file
6-1. Explicit allow, explicit deny (do not use this!, see later text for reasons)
6-2. Only an allow acl-operator
6-3. Corrected example 6-1, explicit deny all
6-4. Example 6-1 once the cache is considered stable
6-5. Using multiple acl Decision Strings per line
6-7. Denying access to a small section of a larger block
6-8. Filtering out unwanted destination sites
6-9. Denying access to sites with the word sex in the URL
6-11. Allowing Web access during the weekend only
6-12. Denying access to FTP sites
6-13. Breaking search site access
6-15. Using ident usernames to deny cache access
6-16. Using Ident to classify users, and using Squid to deny classes
6-17. Using more than one acl operator on an http_access line
6-18. Specifying more than one acl per http_access line
6-19. Logging ident values from specific machines
6-20. Doing ident lookups for unknown machines
6-21. Allowing a subnet range to only get data we already have (hits)
6-22. Using the broken_posts acl-operator
6-23. Using the snmp_community acl type
6-24. Allowing SNMP access from only one machine
6-25. Using the snmp_community acl type
6-26. Limiting download speed by a word in the URL
6-27. Limiting both overall and per-user bandwidth usage
6-28. Using Class 3 Delay Pools
7-1. The cache_peer tag
7-2. The cache_peer_domain tag
7-3. Using acls to select peers
7-4. Passing suspect urls to a filtering cache
7-5. Ignoring Hierarchy Caches for a Local Top-Level Domain
7-6. Bypassing a parent for a local machine
7-7. Changing the Cache Type by Destination Domain
7-9. Sending Queries to a Multicast Server
7-10. Listening for Multicast Queries
7-11. Using CARP Load Factor variables
8-1. Before Accelerator Configuration
8-2. After Accelerator Configuration
8-3. Forwarding Web Requests to a Server on the Same Machine
8-4. Accelerating a Slow Server

Chapter 1. History and Credits

Chapter 2. Installing Squid

Hardware Requirements

Caching stresses certain hardware subsystems more than others. Although the key to good cache performance is good overall system performance, the following list is arranged in order of decreasing importance:

Do not drastically underpower any one subsystem, or performance will suffer. In the case of catastrophic hardware failure you must have a ready supply of alternate parts. When your cache is critical, you should have a (working!) standby machine with operating system and Squid installed. This can be kept ready for nearly instantaneous swap-out. This will, of course, increase your costs, something that you may want to take into account. Chapter 13 covers standby procedures in detail.

Gathering statistics

When deciding on your cache's horsepower, many factors must be taken into account. To decide on your machine, you need an idea of the load that it will need to sustain: the peak number of requests per minute. This number indicates the number of 'objects' downloaded in a minute by clients, and can be used to get an idea of your cache load.

Computing the peak number of requests is difficult, since it depends on the browsing habits of users. This, in turn, makes deciding on the required hardware difficult. If you don't have many statistics as to your Internet usage, it is probably worth your while installing a test cache server (on any machine that you have handy) and pointing some of your staff at it. Using ratios you can estimate the number of requests with a larger user base.

When gathering statistics, make sure that you judge the 'peak' number of requests, rather than an average value. You shouldn't take the number of requests per day and divide, since your peak (during, for example, lunch hour) can be many times your average number of requests.

It's a very good idea to over-estimate hardware requirements. Stay ahead of the growth curve too, since an overloaded cache can spiral out of control due to a transient network problems If a cache cannot deal with incoming requests for some reason (say a DNS outage), it still continues to accept incoming requests, in the hope that it can deal with them. If no requests can be handled, the number of concurrent connections will increase at the rate that new requests arrive.

If your cache runs close to capacity, a temporary glitch can increase the number of concurrent, waiting, requests tremendously. If your cache can't cope with this number of established connections, it may never be able to recover, with current connections never being cleared while it tries to deal with a huge backlog.

Squid 2.0 may be configured to use threads to perform asynchronous Input/Output on operating systems that supports Posix threads. Including async-IO can dramatically reduce your cache latency, allowing you to use a less powerful machine. Unfortunately not all systems support Posix threads correctly, so your choice of hardware can depend on the abilities of your operating system. Your choice of operating system is discussed in the next section - see if your system will support threads there.

Hard Disks

There are numerous things to consider when buying disks. Earlier on we mentioned the importance of disks with a fast random-seek time, and with high sustained-throughput. Having the world's fastest drive is not useful, though, if it holds a tiny amount of data. To cache effectively you need disks that can hold a significant amount of downloaded data, but that are fast enough to not slow your cache to a crawl.

Seek time is one of the most important considerations if your cache is going to be loaded. If you have a look at a disk's documentation there is normally a random seek time figure. The smaller this value the better: it is the average time that the disk's heads take to move from a random track to another (in milliseconds). Operating systems do all sorts of interesting things (which are not covered here) to attempt to speed up disk access times: waiting for disks can slow a machine down dramatically. These operating system features make it difficult to estimate how many requests per second your cache can handle before being slowed by disk access times (rather than by network speed). In the next few paragraphs we ignore operating system readahead, inode update seeks and more: it's a back of the envelope approximation for your use.

If your cache does not use asynchronous Input-Output (described in the Operating system section shortly) then your cache loses a lot of the advantage gained by multiple disks. If your cache is going to be loaded (or is running anywhere approaching capacity according to the formulae below) you must ensure that your operating system supports posix threads!

A cache with one disk has to seek at least once per request (ignoring RAM caching of the disk and inode update times). If you have only one disk, the formula for working out seeks per second (and hence requests per second) is quite simple:

requests per second = 1000/seek time

Squid load-balances writes between multiple cache disks, so if you have more than one data disk your seeks-per-second per disk will be lower. Almost all operating systems will increase random seek time in a semi-linear fashion as you add more disks, though others may have a small performance penalty. If you add more disks to the equation, the requests per second value becomes even more approximate! To simplify things in the meantime, we are going to assume that you use only disks with the same seek time. Our formula thus becomes:

theoretical requests per second =  -----------------
 (seek time)/(number of disks)

Let's consider a less theoretical example: I have three disks - all have 12ms seek times. I can thus (theoretically, as always) handle:

requests per second = 1000/(12/3) = 1000/4 = 250 requests per second

While we are on this topic: many people query the use of IDE disks in caches. IDE disks these days generally have very similar seek times to SCSI disks, and (with DMA-compatible IDE controllers) approach the speed of data transfer without slowing the whole machine down.

Deciding how much disk space to allocate to Squid is difficult. For the pilot project you can simply allocate a few megabytes, but this is unlikely to be useful on a production cache.

The amount of disk space required depends on quite a few factors.

Assume that you were to run a cache just for yourself. If you were to allocate 1 gig of disk, and you browse pages at a rate of 10 megabytes per day, it will take at least 100 days for you to fill the cache.

You can thus see that the rate of incoming cache queries influences the amount of disk to allocate.

If you examine the other end of the scale, where you have 10 megabytes of disk, and 10 incoming queries per second, you will realize that at this rate your disk space will not last very long. Objects are likely to be pushed out of the cache as they arrive, so getting a hit would require two people to be downloading the object at almost exactly the same time. Note that the latter is definitely not impossible, but it happens only occasionally on loaded caches.

The above certainly appears simple, but many people do not extrapolate. The same relationships govern the expulsion of objects from your cache at larger cache store sizes. When deciding on the amount of disk space to allocate, you should determine approximately how much data will pass through the cache each day. If you are unable to determine this, you could simply use your theoretical maximum transfer rate of your line as a basis. A 1mb/s line can transfer about 125000 bytes per second. If all clients were setup to access the cache, disk would be used at about 125k per second, which translates to about 450 megabytes per hour. If the bulk of your traffic is transferred during the day, you are probably transferring 3.6 gigabytes per day. If your line was 100% used, however, you would probably have upgraded it a while ago, so let's assume you transfer 2 gigabytes per day. If you wanted to keep ALL data for a day, you would have to have 2 gigabytes of disk for Squid.

The feasibility of caching depends on two or more users visiting the same page while the object is still on disk. This is quite likely to happen with the large sites (search engines, and the default home pages in respective browsers), but the chances of a user visiting the same obscure page is slim, simply due to the volume of pages. In many cases the obscure pages are on the slowest links, frustrating users. Depending on the number of users requesting pages you should keep pages for longer, so that the chances of different users accessing the same page twice is higher. Determining this value, however, is difficult, since it also depends on the average object size, which, in turn, depends on user habits.

Some people use RAID systems on their caches. This can dramatically increase availability, but a RAID-5 system can reduce disk throughput significantly. If you are really concerned with uptime, you may find a RAID system useful. Since the actual data in the cache store is not vital, though, you may prefer to manually fail-over the cache, simply re-formatting or replacing drives. Sure, your cache may have a lower hit-ratio for a short while, but you can easily balance this minute cost against what hardware to do automatic failover would have cost you.

You should probably base your purchase on the bandwidth description above, and use the data discussed in chapter 11 to decide when to add more disk.

RAM requirements

Squid keeps an in-memory table of objects in RAM. Because of the way that Squid checks if objects are in the file store, fast access to the table is very important. Squid slows down dramatically when parts of the table are in swap.

Since Squid is one large process, swapping is particularly bad. If the operating system has to swap data, Squid is placed on the 'sleeping tasks' queue, and cannot service other established connections. (? hmm. it will actually get woken up straight away. I wonder if this is relevant ?)

Each object stored on disk uses about 75 bytes (? get exact value ?) of RAM in the index. The average size of an object on the Internet is about 13kb, so if you have a gigabyte of disk space you will probably store around about 80 000 objects.

At 75 bytes of RAM per object, 80 000 objects require about six megabytes of RAM. If you have 8gigs of disk you will need 48Mb of RAM just for the object index. It is important to note that this excludes memory for your operating system, the Squid binary, memory for in-transit objects and spare RAM for for disk cache.

So, what should your sustained-thoughput of your disks be? Squid tends to read in small blocks, so throughput is of lesser importance than random seek times. Generally disks with fast seeks are high throughput, and most disks (even IDE disks these days) can transfer data faster than clients can download it from you. Don't blow a year's budget on really high-speed disks, go for lower-seek times instead - or add more disks.

Choosing an Operating System

Where I work, we run many varieties of Unix. When I first installed Squid it was on my desktop Linux machine - if I break it by mistake it's not going to cause users hassles, so I am free to do on it what I wish.

Once I had tested Squid, we decided to allow general access to the cache. I installed Squid on the fastest unused machine we had available at the time: a (then, at least) top of the range Pentium 133 with 128Mb of RAM running FreeBSD.

I was much more familiar with Linux at that stage, and eventually installed Linux on the public cache machine. Though running Linux caused some inconveniences (specifically with low per-process filehandle limits), it was the right choice, simply because I could maintain the machine better. Many times my experience with Linux has gotten me out of potentially sticky situations.

If your choice of operating system saves you time, and runs Squid, use it! Just as I didn't use Digital Unix (Squid is developed on funded Digital Unix machines at NLANR), you don't need to use Linux just because I do.

Most modern operating systems sport both similar performance and similar feature sets. If your system is commonly used and roughly Posix compliant at the source level, it will almost certainly be supported by Squid.

When was the last time you had an outage due to hardware failure? Unless you are particularly unlucky, the interval between hardware failures is low. While the quality of hardware has increased dramatically, software often does not keep pace. Many outages are caused by faulty application of operating system software. You must thus be able to pick up the pieces if your operating system crashes for some reason.

Basic System Setup

Before you even install the operating system, it's best to get an idea as to how the system will look once Squid is up and running. This will allow you to partition the disks on the machine so that their mount path will match Squid's default configuration.

Default Squid directory structure

Normally Squid's directory tree looks like this:


Working through each directory below /usr/local/squid in the order presented above:

bin. The Squid binary and associated tools are stored in this directory. Some tools are included with the Squid source to help you manage and tune your cache server.

cache. Squid has to store cached data on disk somewhere. The path /usr/local/squid/cache is the default location. You can change the location of this directory by editing the Squid config file.

etc. Squid configuration files are stored in this directory. The most commonly changed file in here is squid.conf. We discuss the basic tags in that file in the next chapter.

src. Since you are likely to download the source code for Squid from the net, it is useful to compile the code where you can find it easily. I generally create a src directory and extract the code in there. This way I can revert to a previous version (without downloading it all over again). If you wish, you can easily keep Squid in your /usr/local/src directory, or delete it completely once you have installed the binaries.

Back to the cache directory: if you have more than one partition for the cached data, you can make subdirectories for each of the filesystems in the cache directory. Normally people name these directories cache1, cache2', cache3 and so forth. Your cache directories should be mounted somewhere like /usr/local/squid/cache/1/ and /usr/local/squid/cache/2/. If you have only one cache disk, you can simply name the directory /usr/local/squid/cache/.

In Squid-1.1 cache directories had to be identical in size. This is no longer the case, so if you are upgrading to Squid 2.0 you may be able to resize your cache partitions. To do this, however, you may have to repartition disks and reformat.

When you upgrade to the latest version of Squid, it's a good idea to keep the old working compiled source tree somewhere. If you upgrade to the latest Squid and encounter problems, simply kill Squid, change to the previous source directory and reinstall the old binaries. This is a lot faster than trying to remember which source tree you were running, downloading it, compiling it, applying local patches and then reinstalling.

User and Group IDs

Squid, like most daemon processes on Unix machines, normally runs as the user nobody and with the group nogroup.

For the maximum flexibility in allowing root and non-root users to manipulate the Squid configuration, you should make both a new user and two new groups, specifically for the Squid system, rather than using the nobody and nogroup IDs. Throughout this book we assume that you have done so, and that a group and a user have been created, (both called squid) and a second admin group, called squidadm. The squid user's primary group should be squid, and the user's home directory should be /usr/local/squid (the default squid software install destination).

When you have multiple administrators of a cache machine, it is useful to have a dedicated squidadm group, with sub-administrators added to this group. This way, you don't have to change to the root user whenever you want to make changes to the Squid config. It's possible, for users in the squidadm group to gain root access, so you shouldn't place people without root access in the squidadm group.

When the config file has been changed, a signal has to be sent to the Squid process to inform it that that config files are to be re-read. Sending signals to running processes isn't possible when the signal sender isn't the same userid as the receiver. Other config file maintainers need permission to change their user-id (either by using the 'su' command, or by logging in with another session) to either the root user or to the user Squid is running as.

In some environments cache software maintainers aren't trusted with root access, and the user nobody isn't allowed to log in. The best solution is to allow users that need to make changes to the config file access to a reload script using sudo. Sudo is available for many systems, and source code is available.

In Chapter 4 we go through the process of changing the user-id that Squid runs as, so that files Squid creates are owned by the squid user-id, and by the group squid. Binaries are owned by root, and config files are changeable by the squidadm group.

Getting Squid

Now that your machine is ready for your Squid install, you need to download and install the Squid program. This can be done in two ways: you can download a source version and compile it, or you can download a precompiled binary version and install that, relying on someone else to do the compilation for you.

Binary versions of Squid are generally easier to install than source code versions, specifically if your operating system vendor distributes a package which you can simply install.

Installing Squid from source code is recommended. This method allows you to turn on compile-time options that may not be included in distributed binary versions (one of many examples: SNMP support is not included into the source at compile time unless it is specifically included, and most binary versions available do not include snmp support). If your operating system has been optimized so that Squid can run better (let's say you have increased the number of open filehandles per process) a precompiled binary will not take advantage of this tuning, since your compiler header files are probably different to the ones where the binaries where compiled.

It's also a little worrying running binaries that other people distribute (unless, of course, they are officially supplied by your operating system vendor): what if they have placed a trojan into the binary version? To ensure the security of your system it is recommended that you compile from the official source tree.

Since we suggest installing from source code first, we cover that first: if you have to download a Squid binary from somewhere, simply skip to the next sub-section: Getting a binary version of Squid.

Getting the Squid source code

Squid source is mirrored by numerous sites. For a list of mirrors, have a look at

Deciding which of the available files to download can become an issue, especially if you are not familiar with the Squid version naming convention. Squid is (as of this writing) in version 2. As features are added, the minor version number is incremented (Squid 2.0 becomes Squid 2.1, then Squid 2.2 etc etc). Since new features may introduce new bugs, the first version including new features is distributed as a pre-release (or beta) version. The first pre-release of Squid 1.2 is called squid-2.1.PRE1-src.tar.gz. The second is squid-2.1.PRE2-src.tar.gz. Once Squid is considered stable, a general release version is distributed: the first release version is called squid-2.0.RELEASE-src.tar.gz, the second (which would include minor bugfixes) squid-2.0.RELEASE2-src.tar.gz.

In short, files are named as follows: squid-2.minor-version-number.stability-info.release-number.tar.gz. Unless you are a Squid developer, you should download the last available RELEASE version: you are less likely to encounter bugs this way.

Squid source is normally available via FTP (the File Transfer Protocol), so you should be able to download Squid source by using the ftp program, available on almost every Unix system. If you are not familiar with ftp, you can simply select the mirror closest to you with your browser and save the Squid source to your disk by right-clicking on the filename and selecting save as (do not simply click on the filename - many browsers attempt to extract compressed files, printing the tar file to your browser window: this is definitely not what you want!). Once the download is complete, transfer the file to the cache machine.

Compiling Squid

Compiling Squid is quite easy: you need the right tools to do the job, though. First, let's go through getting the tools, then you can extract the source code package, include optional Squid components (using the configure command) and then actually compile the distributed code into a binary format.

A word of warning, though: this is the stage where most people run into problems. If you haven't compiled source before, try and follow the next section in order - it shouldn't be too bad. If you don't manage to get Squid running, at least you have gained experience.

Compilation options

Squid features are enabled (or disabled) with the configure shell script. Some Squid features have to be specifically enabled when Squid is compiled, which can mean that you have to recompile at a later stage. There are two reasons that a feature can be disabled by default:

You may be wondering why there simply aren't config file options for these less used features. For most of the features there really isn't a reason other than (?minimalisim?). Why have code sitting in the executable that isn't actually used? You can include the features that you might use at some time in the future without detrimental effects (other than a slightly larger binary), so as to avoid having to recompile the Squid source later on.

The configure program also has a second function: with some source code you have to edit a header file which tell the compiler which function calls to use on the system. This very often makes source compilation difficult. With Squid, however, the GNU configure script checks what programs, libraries and function calls are available on your system. This simplifies setup dramatically.

To make configure as generic as possible, it's actually a Bourne Shell /bin/sh script. If you have replaced your /bin/sh shell with a less Posix-capable shell (like ash) you may not be able to run configure. If this is the case you will have to change the first line of the configure script to run the full shell.

all source inclusion options are set with the command './configure option'. On most systems root doesn't have a '.' in their search path for security reasons, so you have to fully specify the path to the binary (hence the '/').

To turn more than one configuration option on at once you simply append each option to the end of the command line. You can, for example, change the prefix install directory and turn Async-IO on with a command like the following (more on what each of these options is for shortly).

 ./configure --prefix=/usr/people/staff/oskar/squid --enable-async-io

Note that only the commonly used configuration options are included here. To get a complete list of options you can run './configure --help'. Many of the resulting options are standard to the GNU configure script that Squid uses, and are used for some things like cross compilation.

If you wish to find out about some of the more obscure options you may have to ask someone on one of the relevant mailing lists, or even read the source code!

Asynchronous IO

Squid 2.0 includes a major performance increase in the form of Async-IO.

It's important to remember that Squid is one processes. In many Internet daemons, more than one copy runs at a time, so if one process is by a system call, it does not effect the other running copies.

Squid is only one process. If the main loop stops running for some reason, all connections are slowed. In all versions of Squid, the main loop uses the select and poll system calls to decide which connections to service. As Squid receives data from the server, it writes the data to disk and to the client.

To write data to disk, a file has to be opened on the cache drive. When lots of clients are opening and closing connections to a busy cache, the main loop has to make lots of calls to open and close network and disk filehandles (note that the word filehandle can refer to both a network connection and an on-disk file). These two functions block the flow of all data through the cache. While waiting for open to return, Squid cannot perform any other functions.

When you enable Async-IO, Squid 2.0 uses threads to open and close filedescriptors. A thread is part of the main Squid program in most ways, except that if it makes use of a blocking system call (such as open), only the thread stops, not the main loop or other threads. Note that there is not one thread per connection.

Using threads to make calls to blocking function calls reduces the latency that a cache adds to each request. (People sometimes worry about the latency that caches add, but if you have a fast enough cache the latency is not an issue - the client sees no noticeable overhead. Network overhead normally outweighs Squid overhead). Async-IO drastically reduces cache overhead when you have a loaded cache.

Unfortunately Posix threads aren't available on all operating systems. This ties your hardware choice into your choice of operating system, since if your operating system does not support threads there may be no choice but to use a faster system, or even to split the load between multiple machines. (? need a table of machines that work ?)

You should probably try and run Squid with Async-IO enabled if you have a few thousand requests per hour. Some systems only support threads properly with a fair amount of initial setup. If your load is low and Async-IO doesn't work straight away you can leave Squid in the default configuration.

Use the --enable-async-io configure option to include the async-io code into Squid.

Running configure

Now that you have decided which options to use, it's time to run configure. Here's an example:

 ./configure --enable-err-language=Bulgarian --prefix=/usr/local

Running ./configure with the options that you have chosen should go smoothly. In the unlikely event that configure returns with an error message, here are some suggestions that may help.

Chapter 3. Squid Configuration Basics

The first high-performance proxy-cache program was developed as part of the Harvest project. The Harvest project was an NSF (?check this info for accuracy?) funded project to create a web indexing system. Part of this project included writing a high-performance cache daemon, or cached (pronounced "Cache-Dee") to speed the re-indexing of pages. Once the project was completed the cached source code was used as the basis for many commercial cache servers, as the source was freely available. Many of the cached developers moved on to or formed companies that developed commercial cache software.

I remember first installing cached: I was boggled at the number of options in the configuration file. I tried working through the options from top to bottom, deciding which to change and which to leave. I had no idea what they all meant. As I worked though the file, I figured more and more options out, though others remained mysteries.

After a lot of changes I tried to start cached, and had no luck. It spat out loads of errors, and I couldn't connect to the machine with my web browser at all. I had no idea what the real problem was - and I changed more and more options with time. This simply buried the real problem beneath hundreds of other possible problems.

Though Squid is now easier to install, the lessons I learned then are still relevant. The default configuration file is probably right for 90% of installations - once you have Squid running, you should change the configuration file one option at a time. Don't get over-ambitious in your changes quite yet! Leave things like refresh rules until you have experimented with the basic options - what port you want your to accept requests on, what user to run as, and where to keep cached pages on your drives.

So that you can get Squid running, this chapter works through the basic Squid options, giving you background information and introducing you to some of the basic concepts. In later chapters you'll move on to more advanced topics.

The Squid config file is not arranged in the order as this book. The config file also does not progress from basic to advanced config options in any specific order, but instead consists of related sections, with all hierarchy settings in a specific section of the file, all access controls in another and so forth.

To make changes detailed in this chapter you are going to have to skip around in the config file a bit. It's probably easiest to simply search for the options discussed in each subsection of this chapter, but if you have some time it will be best if you read through the config file, so that you have an idea of how sections fit together.

The chapter also points out options that may have to be changed on the other 10% of machines. If you have a firewall, for example, you will almost certainly have to configure Squid differently to someone that doesn't.

Version Control Systems

I recommend that you put all Squid configuration files and startup scripts under revision control. If you are like me, you love to play with new software. You change an option, get the program to re-read the configuration file, and see what difference it makes. By repeating this process, I learn what each option does, and at the same time I gain experience, and discover why the program is written the way it is. Quite often configuration files make no sense until you discover the overall structure of the underlying program.

The best way for you to understand each of the options in the Squid config file (and to understand Squid itself) is to experiment with the multitude of options. At some stage in the experimentation stage, you will find that you break something. It's useful to be able to revert to a previous version (or simply to be reminded what changes you have made).

Many readers will already have used a Revision Control System. The RCS system is included with many Unix systems, and source is freely available. For the few that haven't used RCS, however, it's worth including some pointers to some manual pages:


One of the wonders of Unix is the ability to create scripts which reduce the number of commands that you have to type to get something done. I have a short script on all the machines I maintain called rvi. Using rvi instead of vi allows me to use one command to edit files under RCS (as opposed to the customary four). Put this file somewhere in your path and make it executable chmod +x rvi. You can then simply use a command like rvi squid.conf to edit files that are under revision control. This is a lot quicker than running each of the co, rcsdiff and ci commands.

co -l $1
rcsdiff -u $1
ci -u $1

Setting Squid's HTTP Port

The first option in the squid.conf file sets the HTTP port(s) that Squid will listen to for incoming requests.

Network services listen on particular ports. Ports below 1024 can only be used by the system administrator, and are used by programs that provide basic Internet services: SMTP, POP, DNS and HTTP (web). Ports above 1024 are used for untrusted services (where a service does not run as administrator), and for transient connections, such as outgoing data requests.

Typically, web servers listen for incoming web requests (using the HyperText Transfer Protocol - HTTP) on port 80.

Squid's default HTTP port is 3129. Many people run their cache servers on a port which is easier to remember: something like 80 or 8080). If you choose a low-numbered port, you will have to start Squid as root (otherwise you are considered untrusted, and you will not be able to start Squid. Many ISPs use port 8080, making it an accepted pseudo-standard.

If you wish, you can use multiple ports appending a second port number to the http_port variable. Here is an example:

http_port 3128 8080

It is very important to refer to your cache server with a generic DNS name. Simply because you only have one server now does not mean that you should not plan for the future. It is a good idea to setup a DNS hostname for your proxy server. Do this right away! A simple DNS entry can save many hours further down the line. Configuring client machines to access the cache server by IP address is asking for a long, painful transition down the road. Generally people add a hostname like to the DNS. Other people prefer the name proxy, and create a name like

Using Port 80

HTTP defines the format of both the request for information and the format of the server response. The basic aspects of the protocol are quite straight forward: a client (such as your browser) connects to port 80 and asks for the file by supplying the full path and filename that it wishes to download. The client also specifies the version of the HTTP protocol it wishes to use for the retrieval.

With a proxy request the format is only a little different. The client specifies the whole URL instead of just the path to the file. The proxy server then connects to the web server specified in the URL, and sends a normal HTTP request for the page. (? The format of HTTP requests is described in more detail in chapter 4, where you type in an HTTP request, just as a browser would send it to test that the cache is responding to requests - may use the 'client' program instead.?)

Since the format of proxy requests is so similar to a normal HTTP request, it is not especially surprising that many web servers can function as proxy servers too. Changing a web server program to function as a proxy normally involves comparatively small changes to the code, especially if the code is written in a modular manner - as is the Apache web server. In many cases the resulting server is not as fast, or as configurable, as a dedicated cache server can be.

The CERN web server httpd was the first widely available web proxy server. The whole WWW system was initially created to give people easy access to CERN data, and CERN HTTPD was thus the de-facto test-bed for new additions to the initial informal HTTP specification. Most (and certainly at one stage all) of the early web sites ran the CERN server. Many system administrators who wanted a proxy server simply used their standard CERN web server (listening on port 80) as their proxy server, since it could function as one. It is easy for the web server to distinguish a web site request from a normal web page request, since it simply has to check if the full URL is given instead of simply a path name. Given the choice (even today) many system administrators would choose port 80 as their proxy server port simply as 'port 80 is the standard port for web requests'.

There are, however, good reasons for you to choose a port other than 80.

Running both services on the same port meant that if the system administrator wanted to install a different web server package (for extra features available in the new software) they would be limited to software that could perform both as a web server and as a proxy. Similarly, if the same sysadmin found that their web server's low-end proxy module could not handle the load of their ever-expanding local client base, they would be restricted to a proxy server that could function as a web server. The only other alternative is to re-configure all the clients, which normally involves spending a few days apologizing to users and helping them through the steps involved in changing over.

Microsoft use the Microsoft web server (IIS) as a basis for their proxy server component, and Microsoft proxy thus only (? tried once - let's see if it's changed since ?) accepts incoming proxy request on port 80. If you are installing a Squid system to replace either CERN, Apache or IIS running in both web-server and cache-server modes on the same port, you will have to set http_port to 80. Squid is written only as a high-performance proxy server, so there is no way for it to function as a web server, since Squid has no support for reading files from a local disk, running CGI scripts and so forth. There is, however, a workaround.

If you have both services running on the same port, and you cannot change your client PC's, do not despair. Squid can accept requests in web-server format and forward them to another server. If you have only one machine, and you can get your web server software to accept incoming requests on a non-default port (for example 81), Squid can be configured to forward incoming web requests to that port. This is called accelerator mode (since it's initial purpose was to speed up very slow web servers). Squid effectively does some translation on the original request, and then simply acts as if the request were a proxy request and connects to the host: the fact that it's not a remote host is irrelevant. Accelerator mode is discussed in more detail in chapter 9. Until then, get Squid installed and running on another port, and work your way through the first couple of chapters of this book, until you have a working pilot-phase system. Once Squid is stable and tested you can move on to changing web server settings. If you feel adventurous, however, you can skip there shortly!

Where to Store Cached Data

Cached Data has to be kept somewhere. In the section on hardware sizing, we discussed the size and number of drives to use for caching. Squid cannot autodetect where to store this data, though, so you need to let Squid know which directories it can use for data storage.

The cache_dir operator in the squid.conf file is used to configure specific storage areas. If you use more than one disk for cached data, you may need more than one mount point (for example /usr/local/squid/cache1 for the first disk, /usr/local/squid/cache2 for the second). Squid allows you to have more than one cache_dir option in your config file.

Let's consider only one cache_dir entry in the meantime. Here I am using the default values from the standard squid.conf.

cache_dir /usr/local/squid/cache/ 100 16 256 

The first option to the cache_dir tag sets the directory where data will be stored. The prefix value simply has /cache/ tagged onto the end and it's used as the default directory. This directory is also made by the make install command that we used earlier.

The next option to cache_dir is straight forward: it's a size value. Squid will store up to that amount of data in that directory. The value is in megabytes, so of the cache store. The default is 100 megabytes.

The other two options are more complex: they set the number of subdirectories (first and second tier) to create in this directory. Squid makes lots of directories and stores a few files in each of them in an attempt to speed up disk access (finding the correct entry in a directory with one million files in it is not efficient: it's better to split the files up into lots of smaller sets of files... don't worry too much about this for the moment). I suggest that you use the default values for these options in the mean time: if you have a very large cache store you may want to increase these values, but this is covered in the section on

Effective User and Group ID

Squid can only bind to low numbered ports (such as port 80) if it is started as root. Squid is normally started by your system's rc scripts when the machine boots. Since these scripts run as root, Squid is started as root at bootup time.

Once Squid has been started, however, there is no need to run it as root. Good security practice is to run programs as root only when it's absolutely necessary, and for this reason Squid changes user and group ID's once it has bound to the incoming network port.

The cache_effective_user and cache_effective_group tags tell Squid what ID's to change to. The Unix security system would be useless if it allowed all users to change their ID's at will, so Squid only attempts to change ID's if the main program is started as root.

If you do not have root access to the machine, and are thus not starting Squid as root, you can simply leave this option commented out. Squid will then run with whatever user ID starts the actual Squid binary.

As discussed in chapter 2, this book assumes that you have created both a squid user and a squid group on your cache machine. The above tags should thus both be set to "squid".

Access Control Lists and Access Control Operators

Squid could not be used in an ISP environment without a sophisticated access control system. Indeed, Squid should not be used in ANY environment without some kind of basic authentication system. It is amazing how fast other Internet users will find out that they can relay requests through your cache, and then proceed to do so.

Why? Sometimes to obfusticate their real identity, and other times since they have a fast line to you, but a slow line to the remainder of the Internet.

Simple Access Control

In many cases only the most basic level of access control is needed. If you have a small network, and do not wish to use things like user/password authentication or blocking by destination domain, you may find that this small section is sufficient for all your access control setup. If not, you should read chapter 6, where access control is discussed in detail.

The simplest way of restricting access is to only allow IPs that are on your network. If you wish to implement different access control, it's suggested that you put this in place later, after Squid is running. In the meantime, set it up, but only allow access from your PC's IP address.

Example access control entries are included in the default squid.conf. The included entries should help you avoid some of the more obscure problems, such as bandwidth-chewing loops, cache tunneling with SSL CONNECTs and other strange access problems. In chapter 6 we work through the config file's default config options, since some of them are pretty complex.

Access control is done on a per-protocol basis: when Squid accepts an HTTP request, the list of HTTP controls is checked. Similarly, when an ICP request is accepted, the ICP list is checked before a reply is sent.

Assume that you have a list of IP addresses that are to have access to your cache. If you want them to be able to access your cache with both HTTP and ICP, you would have to enter the list of IP addresses twice: you would have lines something like this:

Rule sets like the above are great for small organisations: they are straight forward.

For large organizations, though, things are more convenient if you can create classes of users. You can then allow or deny classes of users in more complex relationships. Let's look at an example like this, where we duplicate the above example with classes of users:

Sure, it's more complex for this example. The benefits only become apparent if you have large access lists, or when you want to integrate refresh-times (which control how long objects are kept) and the sources of incoming requests. I am getting quite far ahead of myself, though, so let's skip back.

We need some terminology to discuss access control lists, otherwise this could become a rather long chapter. So: lines beginning with acl are (appropriately, I believe) acl lines. The lines that use these acls (such as http_access and icp_access in the above example) are called acl-operators. An acl-operator can either allow or deny a request.

So, to recap: acls are used to define classes. When Squid accepts a request it checks the list of acl-operators specific to the type of request: an HTTP request causes the http_access lines to be checked; an ICP request checks the icp_access lists.

Acl-operators are checked in the order that they occur in the file (ie from top to bottom). The frst acl-operator line that matches causes Squid to drop out of the acl list. Squid will not check through all acl-operators if the first denies the request.

In the previous example, we used a src acl: this checks that the source of the request is within the given IP range. The src acl-type accepts IP address lists in many formats, though we used the subnet/netmask in the earlier example. CIDR (Classless Internet Domain Routing) notation can also be used here. Here is an example of the same address range in either notation:

Access control lists inherit permissions when there is no matching acl If all acl-operators in the file are checked, and no match is found, the last acl-operator checked determines whether the request is allowed or denied. This can be confusing, so it's normally a good idea to place a final "catch-all" acl-operator at the end of the list. The simplest way to create such an operator is to create an acl that matches any IP address. This is done with a src acl with a netmask of all 0's. When the netmask arithmetic is done, Squid will find that any IP matches this acl.

Your cache server may well be on the network placed in the relevant allow lists on your cache, and if you were thus to run the client on the cache machine (as opposed to another machine somewhere on your network) the above acl and http_access rules would allow you to test the cache. In many cases, however, a program running on the cache server will end up connecting to (and from) the address '' (also known as localhost). Your cache should thus allow requests to come from the address In the below example we don't allow icp requests from the localhost address, since there is no reason to run two caches on the same machine.

The squid.conf file that comes with Squid includes acls that deny all HTTP requests. To use your cache, you need to explicitly allow incoming requests from the appropriate range. The squid.conf file includes text that reads:


To allow your client machines access, you need to add rules similar to the below in this space. The default access-control rules stop people exploiting your cache, it's best to leave them in.

Ensuring Direct Access to Internal Machines

Acl-operator lines are not only used for authentication. In an earlier section we discussed communication with other cache servers. Acl lines are used to ensure that requests for specific URLs are handled by your cache, not passed on to another (further away) cache.

If you don't have a parent cache (a firewall, or you have a parent ISP cache) you can probably skip this section.

Let's assume that you connect to your ISP's cache server as a parent. A client machine (on your local network) connects to your cache and requests http://www.yourdomain.example/. Your cache server will look in the local cache store. If the page is not there, Squid wil will connect to it's configured parent (your ISP's cache: across your serial link), and request the page from there. The problem, though, is that there is no need to connect across your internet line: the web server is sitting a few feet from your cache in the machine room.

Squid cannot know that it's being very inefficient unless you give it a list of sites that are "near by". This is not the only way around this problem though: your browser could be configure to ignore the cache for certain IPs and domains, and the request will never reach the cache in the first place. Browser config is covered in Chapter 5, but in the meantime here is some info on how to configure Squid to communicate directly with internal machines.

The acl-operators always_direct and never_direct determine whether to pass the connection to a parent or to proceed directly.

The following is a set of operators are based on the final configuration created in the previous section, but using never_direct and always_direct operators. It is assumed that all servers that you wish to connect to directly are in the address ranges specified in with the my-iplist directives. In some cases you may run a web server on the same machine as the cache server, and the localhost acl is thus also considered local.

The always_direct and never_direct tags are covered in more detail in Chapter 7, where we cover hierarchies in detail.

Squid always attempts to cache pages. If you have a large Intranet system, it's a waste of cache store disk space to cache your Intranet. Controlling which URLs and IP ranges not to cache are covered in detail in chapter 6, using the no_cache acl operator.

Communicating with other proxy servers

Squid supports the concept of a hierarchy of proxies. If your proxy does not have an object on disk, it's default action is to connect to the origin web server and retrieve the page. In a hierarchy, your proxy can communicate with other proxies (in the hope that one of these servers will have the relevant page). You will, obviously, only peer with servers that are 'close' to you, otherwise you would end up slowing down access. If access to the origin server is faster than access to neighboring cache servers it is not a good idea to get the page from the slower link!

Having the ability to treat other caches as siblings is very useful in some interactions. For example: if you often do business with another company, and have a permanent link to their premises, you can configure your cache to communicate with their cache. This will reduce overall latency: it's almost certainly faster to get the page from them than from the other side of the country.

When querying more than one cache, Squid does not query each in turn, and wait for a reply from the first before querying the second (since this would create a linear slowdown as you add more siblings, and if the first server stops responding, you would slow down all incoming requests). Squid thus sends all ICP queries together - without waiting for replies. Squid then puts the client's request on hold until the first positive reply from a sibling cache is received, and will retrieve the object from the fastest-replying cache server. Since the earliest returning reply packet is usually on the fastest link (and from the least loaded sibling server), your server gets the page fast.

Squid will always get the page from the fastest-responding cache - be it a parent or a sibling.

The cache_peer option allows you to specify proxy servers that your server is to communicate with. The first line of the following example configures Squid to query the cache machine cache.myparent.example as a parent. Squid will communicate with the parent on HTTP port 3128, and will use ICP to query the server using port 3130. Configuring Squid to query more than one server is easy: simply add another cache_peer line. The second line configures cache.sibling.example as a sibling, listening for HTTP request on port 8080 and ICP queries on port 3130.

cache_peer cache.myparent.example parent 3128 3130
cache_peer cache.sibling.example sibling 8080 3130

If you do not wish to query any other caches, simply leave all cache_peer lines commented out: the default is to talk directly to origin servers.

Cache peering and hierarchy interactions are discussed in quite some detail in this book. In some cases hierarchy setups are the most difficult part of your cache setup process (especially in a distributed environment like a nationwide ISP). In depth discussion of hierarchies is beyond the scope of this chapter, so much more information is given in chapter 8. There are cases, where you need at least one hierarchy line to get Squid to work at all. This section covers the basics, just for those setups.

You only need to read this material if one of the following scenarios applies to you:

Firewall Interactions

Firewalls can make cache configuration hairy. Inter-cache protocols generally use packets which firewalls inherently distrust. Most caches (Squid included) use ICP, which is a layer on top of UDP. UDP is difficult to make secure, and firewall administrators generally disable it if at all possible.

It's suggested that you place your cache server on your DMZ (if you have one). There are a few advantages to this:

The remainder of this section should help you getting Squid and your firewall to co-operate. A few cases are covered for each type of firewall: the cache inside the firewall; the cache outside the firewall; and, finally, on the DMZ.

Proxying Firewalls

The vast majority of firewalls no nothing about ICP. If, on the other hand, your firewall does not support HTTP, it's a good time to have a serious talk to the buyer that had an all-expenses-paid weekend on the firewall supplier. Configuring the firewall to understand ICP is likely to be painful, but HTTP should be easy.

If you are using a proxy-level firewall, your client machines are probably configured to use the firewall's internal IP address as their proxy server. Your firewall could also be running in transparent mode, where it automatically picks up outgoing web requests. If you have a fair number of client machines, you may not relish the idea of reconfiguring all of them. If you fall into this category, you may wish to put your firewall on the outside (or on the DMZ) and configure the firewall to pass requests to the cache, rather than reconfiguring all client machines.


The cache is considered a trusted host, and is protected by the firewall. You will configure client machines to use the cache server in their browser proxy settings, and when a request is made, the cache server will pass the outgoing request to the firewall, treating the firewall as a parent proxy server. The firewall will then, connect to the destination server. If you have a large number of clients configured to use the firewall as their proxy server, you could get the firewall to hand-off incoming HTTP requests back into the network, to the cache server. This is less efficient though, since the cache will then have to re-pass these requests through the firewall to get to the outside, using the parent option to cache_peer. Since the latter involves traffic passing through the firewall twice, your load is very likely to increase. You should also beware of loops, with the cache server parenting to the firewall and the firewall handing-off the cache's request back to the cache!

As described in chapter 1, Squid will also send ICP queries to parents. Firewalls don't care for UDP packets, and normally log (and then discard) such packets.

When Squid does not receive a response from a configured parent, it will mark the parent as down, and proceed to go directly.

Whenever Squid is setup to use a parent that does not support ICP, the cache_peer line should include the "default" and "no-query" options. These options stop Squid from attempting to go direct when all caches are considered down, and specify that Squid is not to send ICP requests to that parent.

Here is an example config entry:

cache_peer inside.fw.address.domain parent 3128 3130 default no-query


There are only two major reasons for you to put your cache outside the firewall:

One: Although squid can be configured to do authentication, this can lead to the duplication of effort (you will encounter the "add new staff to 500 servers" syndrome). If you want to continue to authenticate users on the firewall, you will have to put your cache on the outside or on the DMZ. The firewall will thus accept requests from clients, authenticate them, and then pass them on to the cache server.

Two: Communicating with cache hierarchies is easy. The cache server can communicate with other systems using any protocol. Sibling caches, for example, are difficult to contact through a proxying firewall.

You can only place your cache outside if your firewall supports hand-offs. Browsers inside will connect to the firewall and request a URL, and the firewall will connect to the outside cache and request the page.

If you place your cache outside your firewall, you may find that your client PC's have problems connecting to internal web servers (your intranet, for example, may be unreachable). The problem is that the cache is unable to connect back through to your internal network (which is actually a good thing: don't change that). The best thing to do here is to add exclusions to your browser settings: this is described in Chapter 5 - you should specifically have a look at the section on browser autoconfig. In the meantime, let's just get Squid going, and we will configure browsers once you have a cache to talk to.

Since the cache is not protected by the firewall, it must be very carefully configured - it must only accept requests from the firewall, and must not run any strange services. If possible, you should disable telnet, and use something like SSH (Secure SHell) instead. The access control lists (which you will setup shortly) must only allow the firewall, otherwise people will be able to relay their requests through your cache, using your bandwidth.

If you place the cache outside the firewall, you client PC's will be configured to use the firewall as their proxy server (this is probably the case already). The firewall must be configured to hand-off client HTTP requests to the cache server. The cache must be configured to only allow HTTP requests when from the firewall's outside IP address. If not configured this way, other Internet users could use your cache server as a relay, using your bandwidth and hardware resources for illegitimate (and possibly illegal) purposes.

With your cache server on the outside network, you should treat the machine as a completely untrusted host, lest a cracker find a hole somewhere on the system. It is recommended that you place the cache server on a dedicated firewall network card, or on a switched ethernet port. This way, if your cache server were to be cracked, the cracker would only be able to read passing HTTP data. Since the majority of sensitive information is sent via email, this would reduce the potential for sensitive data loss.

Since your cache server only accepts requests from the firewall, there is no cache_peer line needed in the squid.conf. If you have to talk to your ISP's cache you will, of course, need one: see the section on this a bit further back.


The best place for a cache is your DMZ.

If you are concerned with the security of your cache server, and want to be able to communicate with outside cache servers (using ICP), you may want to put your cache on the DMZ.

With Squid on your DMZ, internal client PCs are setup to proxy to the firewall. The firewall is then responsible for handing-off these HTTP requests to the cache server (so the firewall in fact treats the cache server as a parent).

Since your cache server is (essentially) on the outside of the firewall, the cache doesn't need to treat the firewall as a parent or sibling: it only accepts requests from the firewall: it never passes them to the firewall.

If your cache is outside your firewall, you will need to configure your client PC's not to use the firewall as a proxy server for internal hosts. This is quite easy, and is discussed in the chapter on browser configuration.

Since the firewall is acting as a filter between your cache and the outside world, you are going to have to open up some ports on the firewall. The cache will need to be able to connect to port 80 on any machine on the outside world. Since some valid web servers will run on ports other than 80, you should consider allowing connections to any port from the cache server. In short, allow connections to:

If you are going to communicate with a cache server outside the firewall, you will need even more ports opened. If you are going to communicate with ICP, you will need to allow UDP traffic from and to your cache machine on port 3130. You may find that the cache server that you are peering with uses different ports for reply packets. It's probably a bad idea to open all UDP traffic, though.

Packet Filtering firewalls

Squid will normally live on the inside of your packet-filtering firewall. If you have a DMZ, it may be best to put your cache on this network, as you may want to allow UDP traffic to and from the cache server (to communicate with other caches).

To configure your firewall correctly, you should make the minimum number of holes in your filter set. In the remainder of this section we assume that your internal machines can connect to the cache server unimpeded. If your cache is on the DMZ (or outside the firewall altogether) you will need to allow TCP connections from your internal network (on a random source port) to the HTTP port that Squid will be accepting requests on (this is the port that you set a bit earlier, in the "Setting Squid's HTTP Port" section of this chapter.

First, let's consider the firewall setup when you do not query any outside caches. On accepting a request, Squid will attempt to connect to a machine on the Internet at large. Almost always, the destination port will be the default HTTP port, port 80. A few percent of the time, however, the request will be destined for a high-numbered port (any port number higher than 1023 is a high-numbered port). Squid always sources TCP requests from a high-numbered port, so you will thus need to allow TCP requests (all HTTP is TCP-based) from a random high-numbered port to both port 80 and any high-numbered port.

There is another low-numbered port that you will probably need to open. The HTTPS port (used for secure Internet transactions) is normally listening on TCP port 443, so this should also be opened.

In the second situation, let's look at cache-peering. If you are planning to interact with other caches, you will need to open a few more ports. First, let's look at ICP. As mentioned previously, ICP is UDP-based. Almost all ICP-compliant caches listen for ICP requests on UDP port 3130. Squid will always source requests from port 3130 too, though other ICP-compliant caches may source their requests from a different port.

It's probably not a good idea to allow these UDP packets no matter what source address they come from. Your filter should probably specify the IP addresses for each of the caches that you wish to peer from, rather than allowing UDP packets from any source address. That should be it: You should now be able to save the config file, and get ready to start the Squid program.

Chapter 4. Starting Squid

Before Running Squid

Before we can start Squid, we have to create a few directories on the system. It's important that these directories have the correct permissions, otherwise someone with a login on the cache may be able to gain root access. Let's work through the default directory tree, and set the permissions on each directory correctly. Since you may have special requirements, I won't simply give you a sequence of commands to run: if you need to use different permissions, it's important to understand the possible consequences.

Subdirectory Permissions

In Chapter 2 we created a squid user and group, and created another group, squidadm for the people that will maintain the cache. When Squid starts up, it changes it's user and group ids to squid (thanks to the cache_effective_user and cache_effective_group tags in squid.conf.) Changing userids reduces the chance of a complete exploit because of a bug in Squid. It's important, however, to remember that users in the squidadm group can probably get root on your machine, so you should not put people that do not already have root on the machine in that group: it's just so that you don't have to su to root continuously.

Walking the Directory Tree

Let's start with the /usr/local/squid/ directory. If your system starts Squid at bootup, a startup script running as root starts the program /usr/local/squid/bin/squid. If someone were to replace this binary with a trojan, they could gain root access. The /usr/local/squid/ directory should be owned by root, group root, and should not be writable by anyone else. This stops someone from moving the entire bin directory to (say), and creating a new bin directory which contains their own squid binary. Use the following commands to set the permissions on this directory correctly:

chown root:root /usr/local/squid/
chmod 755 /usr/local/squid/

Since we have already introduced the /usr/local/squid/bin directory, let's set it's permissions correctly next. If the directory itself was writeable by malicious users, we would have the same problem that we described above. Let's change it to be owned by root, group root, and make sure that only these root can write to the directory. We also need the files in this directory to be readable (and executable) by everyone, so that normal users can run programs like client. There are no setuid binaries in this directory, and if the rest of the files have the correct permissions, there is no reason not to let users into this directory.

cd /usr/local/squid/bin
chown root:root .  chown root:root *
chmod 755 . *

Config files all live in the /usr/local/squid/etc/ directory. If a user can write to these files, they can almost certainly do malicious things. Because of this, you should not let normal users edit these files: only users which already have root access should be allowed to edit squid.conf. Earlier in the book, we created a squidadm for these users.

The /usr/local/squid/etc/ directory should be owned by root, group squidadm, so that squid-administrators would be able to create and update config files.

Many of you will not have encountered chown commands which use more than three numbers before. The following command sets the sticky bit on the directory. Let's assume that my primary group-id is staff (not squidadm.) On some systems, any file that I create will be owned by group staff, even if the directory is owned by the squidadm group. On these systems this would be a security problem: if I create the squid.conf file, people in the staff group may be able to make changes to the file.

With the sticky bit set on the directory, any files I create will be owned by the squidadm group. As I have said: this isn't necessary on some operating systems, but these permissions shouldn't have any adverse effect.

cd /usr/local/squid/etc
chmod 2775 .
chown root:squidadm . *

When you use RCS (introduced in Chapter 2), the revision history of a file is stored in an RCS logfile. These files will normally be created in the current directory (the ci command appends a comma to the filename to decide the name of the logfile, leading to filenames like squid.conf,v.) If you don't want your directory cluttered with these files, you can create an RCS directory, and move RCS files into it. The Revision Control System only stores logfiles in the current directory if an RCS directory doesn't exist, if one does, all new log files are created in it.

If someone can gain access to the log files, they essentially have write access to original file, since when you check a file out (to make changes to it) the log file is considered to be the authoritive source. Don't forget to change the permissions on the RCS log files Squid doesn't create an RCS directory automatically; we create it in the example below.

# first, make the RCS directory
cd /usr/local/squid/etc
mkdir RCS
# move any RCS logfiles into the RCS directory, so that they don't
# clutter the config-file directory
mv *,v RCS
# make sure that the RCS directory is owned by the right people, and
# can be writeable by them
chown root:squidadm RCS
chmod 2770 RCS
# change the permissions of the files in the RCS directory to match
# newly created files
chown root:squidadm RCS/*
chmod 770 RCS/*

Cache log files should be confidential. You (and other Squid administrators) may have to look at them occasionally, but other users should have no access to the files. Squid runs as the squid user, though, and needs to create the logs, so any directory we make needs to be writeable by the squid user too.

chown squid:squidadm /usr/local/squid/logs
chmod 770 /usr/local/squid/logs

Object Store Directory Permissions

As you may recall from Chapter 3, downloaded objects are placed in a hierarchy of swap directories. Squid.conf contains a cache_store line for each directory that files are to be stored in, and specifies the number of subdirectories that are to be created in each cache store (most people leave this at the default, 4096 directories per cache store.)

Squid's -z command-line option will create the appropriate cache-swap directories (since creating them by hand would be painful!) If the top-level cache directory specified in squid.conf does not exist, Squid will attempt to create it too, as squid, group squid (or whatever cache_effective_user/cache_effective_group is set to in squid.conf.) Since we changed the permissions on /usr/local/squid above so that only root can write to this directory, Squid's directory create will fail. Instead, let's create these directories and set their permissions manually.

If someone has read access to the cache logs, they can invade people's privacy. It may seem harmless to let people access the cache store indiscriminately, but I contend that it isn't.

Many web accesses reveal something about the person examining the page, be it their sexuality, their financial status, or their job satisfaction. This is why we stop people accessing log files and finding out who went to what pages, right? Well, it also means that we must stop people accessing the cache store directly. In many systems using cryptography, you can discover much about the nature of the contents of the traffic by traffic analysis, relating traffic flows to other events. If people only have access to the cache using a browser, it difficult for them to associate any hits they see with a given person. If they can examine other information about the object (say, the time that the file was created) they may be able to discover information about the person that requested the object. A simple example: let's say that someone connected to a job-search site at the middle of the night. You can immediately narrow down the list of possible requestors to night-staff. If you can find out who was on duty on that day, you narrow the number of possible requestors even more.

Let's change the permissions on the cache store so that only squid-administrators can access files in it. Note that you are going to have to repeat this process for every cache_dir in the squid.conf file.

mkdir /usr/local/squid/cache/
chown squid:squidadm /usr/local/squid/cache/
chmod 770 /usr/local/squid/cache/

Once the permissions on the cache directories are set correctly, you can run squid -z.

Your output should look something like this:

cache1:~ # /usr/local/squid/bin/squid -z
1999/06/12 19:15:34| Creating Swap Directories
cache1:~ #

Running Squid

Squid should now be configured, and the directories should have the correct permissions. We should now be able to start Squid, and you can try and access the cache with a web browser. Squid is normally run by starting the RunCache script. RunCache (as mentioned ealier) restarts Squid if it dies for some reason, but at this stage we are merely testing that it will run properly: we can add it to startup scripts at a later stage.

Programs which handle network requests (such as inetd and sendmail) normally run in the background. They are run at startup, and log any messages to a file (instead of printing it to a screen or terminal, as most user-level programs do.) These programs are often referred to as daemon programs. Squid is such a program: when you run the squid binary, you should be immediately returned to the command line. While it looks as if the program ran and did nothing, it's actually sitting in the background waiting for incoming requests. We want to be able to see that Squid's actually doing something useful, so we increase the debug level (using -d 1) and tell it not to dissapear into the background (using -N.) If your machine is not connected to the Internet (you are doing a trial squid-install on your home machine, for example) you should use the -D flag too, since Squid tries to do DNS lookups for a few common domains, and dies with an error if it is not able to resolve them.

The following output is that printed by a default install of Squid:

cache1:~ # /usr/local/squid/bin/squid -N -d 1 -D

Squid reads the config file, and changes user-id's here:

1999/06/12 19:16:20| Starting Squid Cache version 2.2.DEVEL3 for i586-pc-linux-gnu...
1999/06/12 19:16:20| Process ID 4121

Each concurrent incoming request uses at least one filedescriptor. 256 filedescriptors is only enough for a small, lightly loaded cache server, see Chapter 12 for more details. Most of the following is diagnostic:

1999/06/12 19:16:20| With 256 file descriptors available
1999/06/12 19:16:20| helperOpenServers: Starting 5 'dnsserver' processes
1999/06/12 19:16:20| Unlinkd pipe opened on FD 13
1999/06/12 19:16:20| Swap maxSize 10240 KB, estimated 787 objects
1999/06/12 19:16:20| Target number of buckets: 15
1999/06/12 19:16:20| Using 8192 Store buckets, replacement runs every 10 seconds
1999/06/12 19:16:20| Max Mem  size: 8192 KB
1999/06/12 19:16:20| Max Swap size: 10240 KB
1999/06/12 19:16:20| Rebuilding storage in Cache Dir #0 (DIRTY)

When you connect to an ftp server without a cache, your browser chooses icons to match the files based on their filenames. When you connect through a cache server, it assumes that the page returned will be in html form, and will include tags to load any images so that the directory listing looks normal. Squid adds these tags, and has a collection of icons that it refers clients to. These icons are stored in /usr/local/squid/etc/icons/. If Squid has permission problems here, you need to make sure that these files are owned by the appropriate users (in the previous section we set permissions on the files in this directory.)

1999/06/12 19:16:20| Loaded Icons.

The next few lines are the most important. Once you see the Ready to serve requests line, you should be able to start using the cache server. The HTTP port is where Squid is waiting for browser connections, and should be the same as whatever we set it to in the previous chapter. The ICP port should be 3130, the default, and if you have included other protocols (such as HTCP) you should see them here. If you see permission denied errors here, it's possible that you are trying to bind to a low-numbered port (like 80) as a normal user. Try run the startup command is root, or (if you don't have root access on the machine) choose a high-numbered port. Another common error message at this stage is Address already in use. This occurs when another process is already listening to the given port. This could be because Squid is already started (perhaps you are upgrading from an older version which is being restarted by the RunCache script) or you have some other process listening on the same port (such as a web server.)

1999/06/12 19:16:20| Accepting HTTP connections on port 3128, FD 35.
1999/06/12 19:16:20| Accepting ICP messages on port 3130, FD 36.
1999/06/12 19:16:20| Accepting HTCP messages on port 4827, FD 37.
1999/06/12 19:16:20| Ready to serve requests.

Once Squid is up-and-running, it reads the cache-store. Since we are starting Squid for the first time, you should see only zeros for all the numbers below:

1999/06/12 19:16:20| storeRebuildFromDirectory: DIR #0 done!
1999/06/12 19:16:25| Finished rebuilding storage disk.
1999/06/12 19:16:25|         0 Entries read from previous logfile.
1999/06/12 19:16:25|         0 Entries scanned from swap files.
1999/06/12 19:16:25|         0 Invalid entries.
1999/06/12 19:16:25|         0 With invalid flags.
1999/06/12 19:16:25|         0 Objects loaded.
1999/06/12 19:16:25|         0 Objects expired.
1999/06/12 19:16:25|         0 Objects cancelled.
1999/06/12 19:16:25|         0 Duplicate URLs purged.
1999/06/12 19:16:25|         0 Swapfile clashes avoided.
1999/06/12 19:16:25|   Took 5 seconds (   0.0 objects/sec).
1999/06/12 19:16:25| Beginning Validation Procedure
1999/06/12 19:16:26| storeLateRelease: released 0 objects
1999/06/12 19:16:27|   Completed Validation Procedure
1999/06/12 19:16:27|   Validated 0 Entries
1999/06/12 19:16:27|   store_swap_size = 21k

Testing Squid

If all has gone well, we can begin to test the cache. True browser access is only covered in the next chapter, and there is a whole chapter devoted to configuring your browser. Until then, testing is done with the client program, which is included with the Squid source, and is in the /usr/local/squid/bin directory.

The client program connects to a cache and request a page, and prints out useful timing information. Since client is available on all systems that Squid runs on, and has the same interface on all of them, we use it for the initial testing.

At this stage Squid should be in the foreground, logging everything to your terminal. Since client is a unix program, you need access to a command prompt to run it. At this stage it's probably easiest to simply start another session (this way you can see if errors are printed in the main window).

The client program is compiled to connect to localhost on port 3128 (you can override these defaults from the command line, see the output of client -h for more details.)

If you are running client on the cache server, and are using port 3128 for incoming requests, you should be able to type a command like this, and the client program will retrieve the page through the cache server:


If your cache is running on a different machine you will have to use the -h and -p options. The following command will connect to the machine cache.qualica.comf on port 8080 and retrieve the above web page.

The client program can also be used to access web sites directly. As you may remember from reading Chapter 2, the protocol that clients use to access pages through a cache is part of the HTTP specification. The client program can be used to send both "normal" and "cache" HTTP requests. To check that your cache machine can actually connect to the outside world, it's a good idea to test access to an outside web server.

The next example will retrieve the page at, and send the html contents of the page to your terminal.

If you have a firewall between you and the internet, the request may not work, since the firewall may require authentication (or, if it's a proxy-level firewall and is not doing transparent proxying of the data, you may explicitly have to tell client to connect to the machine.) To test requests through the firewall, look at the next section.

A note about the syntax of the next request: you are telling client to connect directly to the remote site, and request the page /. With a request through a cache server, you connect to the cache (as you would expect) and request a whole url instead of just the path to a file. In essence, both normal-HTTP and cache-HTTP requests are identical; one just happens to refer to a whole URL, the other to a file.

Client can also print out timing information for the download of a page. In this mode, the contents of thi page isn't printed: only the timing information is. The zero in the below example indicates that Squid is to retrieve the page until interrupted (with Control-C or Break.) If you want to retrieve the page a limited number of times, simply replace the zero with a number.

Testing a Cache or Proxy Server with Client

Now that you have client working, you

If the request through the cache returned the same page as you retrieved with direct access (you didn't receive an error message from Squid), Squid should be up and running. Congratulations! If things aren't going so well for you, you will have received an error message here. Normally, this is because of the acls described in the previous chapter. First, you should have a look at the terminal where you are running Squid (Or, if you are skipping ahead and have put Squid in the background, in the /usr/local/squid/logs/cache.log file.) If Squid encountered some sort of problem, there should be an error or warning in this file. If there are no messages here, you should look at the /usr/local/squid/logs/access.log file next. We haven't coverd the details of this file yet, but they are coverded in the next section of this chapter. First, though, let's see if your cache can process requests to internal servers. There are many cases where a request will work to internal servers but not to external machines.

Access.log basics

The access.log file logs all incoming requests. chapter 11 covers the fields in the access.log in detail. The most important fields are the URL (field 7), and hierarchy access type (field 9) fields. Note that a "-" indicates that there is no data for that field.

The following example access.log entries indicate the changes in log output when connecting to another server, without a cache, with a single parent, and with multiple parents.

Though fields are seperated by spaces, fields can contain sub-fields, where a "/" indicates the split.

When connecting directly to a destination server, field 9 contains two subfields - the key word "DIRECT", followed by the name of the server that it is connecting to. Access to local servers (on your network) should always be DIRECT, even if you have a firewall, as discussed in section 3.1.2. The acl operator always_direct controls this behaviour.

905144366.259   1010 TCP_MISS/200 20868 GET - DIRECT/ text/html

When you have configured only one parent cache, the hierarchy access type indicates this, and includes the name of that cache.

905144426.435    289 TCP_MISS/200 20868 GET - SINGLE_PARENT/ text/html

There are many more types that can appear in the hierarchy access information field, but these are covered in chapter 11.

Another useful field is the 'Log Tag' field, field four. In the following example this is the field "TCP_MISS/200".

905225025.225    609 TCP_MISS/200 10089 GET - DIRECT/ text/html

A MISS indicates that the request was already stored in the cache (or that the page contained headers indicating that the page was not to be cached). A HIT would indicate that the page was already stored in the cache. In the latter case the request time for a remote page should be substantially less than the first occurence in the logs.

The time that Squid took to service the request is the second field. This value is in milliseconds. This value should approach that returned by examining a client request, but given operating system buffering there is likely to be a discrepancy.

The fifth field is the size of the page returned to the client. Note that an aborted request can end up downloading more than this from the origin server if the quick_abort feature set is turned on in the Squid config file.

Here is an example request direct from the origin server:

905230201.136   6642 TCP_MISS/200 20847 GET - DIRECT/ text/html

If we use client to fetch the page a short time later, a HIT is returned, and the time is reduced hugely.

905230209.899    151 TCP_HIT/200 20869 GET - NONE/- text/html

Some of you will have noticed that the size of the hit has increased slightly. If you have checked the size of a request from the origin server and compared it to that of the same page through the cache, you will also note that the size of the returned data has increased very slightly. Extra headers are added to pages passing through the cache, indicating which peer the page was returned from (if applicable), age information and other information. Clients never see this information, but it can be useful for debugging.

Since Squid 1.2 has support for HTTP/1.1, extra features can be used by clients accessing a copy of a page that Squid already has. Certain extra headers are included into the HTTP headers returned in HITS, indicating support for features which are not available to clients when returning MISSes. In the above example Squid has included a header in the page indicating that range-request are supported.

If Squid is performing correctly, you should shut Squid down and add it to your startup files.

Since Squid maintains an in-memory index of all objects in the cache, a kill -9 could cause corruption, and should never be used. The correct way to shutdown Squid is to use the command:

	cache1:~ # ~squid/bin/squid -k shutdown

Squid command-line options are covered in chapter 10.

Chapter 5. Browser Configuration


Squid is the server half of a client-server relationship. Though you have configured Squid, your client (the browser) is still configured to talk to the menagerie of servers that make up the Internet.

You have already used the client program included with Squid to test that the cache is working. Browsers are more complicated to configure than client, especially since there are so many different types of browser.

This chapter covers the three most common browsers. It also includes information on the proxy configuration of Unix tools, since you may wish to use these for automatic download of pages. Once your browser is configured, some of the proxy-oriented features of browsers are covered. Many browsers allow you to force your cache server to reload the page, and have other proxy-specific features.

So that you can skip sections in this chapter that you don't need to read, browsers are configured in the following order: Netscape Communicator, Microsoft Internet Explorer, Opera and finally Unix Clients.

You can configure most browsers in more than one way. The first method is the simplest for a sysadmin, the second is simplest for the user. Since this book is written for system administrators, we term the first basic configuration, the second advanced configuration.

Host name

It's very important to use a proxy specific host name. If you decide to move the cache to another machine at a later stage you will find that it's much easier to change DNS settings than to change the configuration of every browser on your network.

If your operating system supports IP aliases you should organize a dedicated IP address for the cache server, and use the tcp_incoming_address and tcp_outgoing_address squid.conf options to make Squid only accept incoming HTTP requests on that IP address.

There isn't really a naming convention for caches, but people generally use one of the following: cache, proxy, www-proxy, www-cache, or even the name of the product they are using; squid, netapp, netscape. Some people also include the location of the cache, and configure people in a region to talk to their local cache. More and more people are simply using cache, and it's the suggested name. If you wish to use regional names, you can use something along the lines of region.cache.domain.example.

Your choice of port has already been discussed. Have a look at HTTP:port in the index for more information.

Internet Explorer 4.0

Select the View menu option Select Internet Options Click on the Connection tab Select Access the Internet using a proxy server Type in your hostname in the Address: field, and your chosen port in the Port: field. Internet Explorer can attempt to connect directly to the destination server if the URL you are going to is in the local domain (? I presume ?). You should turn this option on, so that local accesses are not cached, and do not pass through the cache server. If you have more than one domain, you will have to specifically change options so that all your domains are ignored, using the Advanced button.

In the advanced menu, you can configure per-protocol cache server/port pairs, or you can type in only the first proxy/port pair, and select Use the same proxy for all protocols. Although Squid doesn't normally work with SOCKS, it's rarely used, so you can probably use the same proxy for all protocols.

The main advantage of using the Advanced menu is the ability to specify which domains are to be connected to directly, rather than through the proxy server. If all your local sites' hostnames begin with intranet, you can simply put that into the box titled Do not use proxy for addresses beginning with. You can add more than one exception by using a semicolon (;) between entries.

You will probably wish to exclude all local sites too. Since the exception list allows you to use a * character for what is known as a wildcard match, you can add *.localdomain.example, and all hosts in your domain will be accessed directly. Many people access local sites by IP address, rather than by name. Since the exception list matches against the URL (??) these will still pass through the cache, and you will need to add an IP address range to the list of hosts to exclude: 192.168.0.* should do nicely.

To reduce the local browser cache space (as discussed in the Netscape section in the previous section):

In the Temporary Internet files section, click the Settings
Move the slider all the way to the left.

Since Squid-2.0 and above handle HTTP/1.1 correctly, you should also configure Internet Explorer to use HTTP/1.1 when communicating with the proxy server:

Internet Options
Advanced tab
Scroll down until you see HTTP 1.1 Settings
Tick Use HTTP 1.1 through proxy server

(? I believe that opera is the third most common browser ?) (? I don't have a machine with it on... since I run Linux?)

Testing the Cache

As you can see, pressing reload in Netscape (and some other browsers) doesn't simply re-fetch the page, it forces the cache not to serve the cached page. Many people doing tests of how the cache increases performance simply press reload, and believe that there has been no change in speed. The cache is, in fact, re-downloading the page from the origin server, so a speed increase is impossible.

To test the cache properly you need two machines setup to access the cache, and a page that does not contain do not cache me headers. Pages that use ASP often include headers that force Squid not to cache the page, even if the authors are not aware of it's implications.

So, to test the cache, choose a site that is off your local network (for a marked change, choose one in a different country) and access it from the first machine. Once it has download, change to the second machine and re-download the page. Once the page has downloaded there, check that the page is marked as a 'HIT' (in the file called access.log - the basics of which are covered earlier in this book). If the second accesses were marked as misses, it is probably because the origin server is asking Squid not to cache the page. Try a different page and see difference the cache makes to browsing speed.

Many people are looking for an increase in performance on problem pages, since this is when people believe that they are getting the short end of the stick. If you choose a site that is too close, you may only be able to see a difference in the speed in the transaction-time field of the access.log.

Since you have a completely unloaded cache, you should access a local, unloaded web server a few times, and see what kind of latency you experience with the cache. If you have time, print out some of the access log entries. If, some time in the future, you are unsure as to the cache load, you can compare the latency added now to the latency added by the same cache later on; if there is a difference you know it's time to upgrade the cache.

Cache Auto-config

Client browsers can have all options configured manually, or they can be configured to download a autoconfig file (every time the startup), which provides all of the information about your cache setup.

Each URL referenced (be it the URL that you typed, or the URL for a graphic on the page yet to be retrieved) is checked against the list of rules. You should keep the list of rules as short as possible, otherwise you could end up slowing down page loads - not at the cache level, but at the browser.

Autoconfig Script Coding

The autoconfig file is actually a Java function, put in a file and served by your standard web server program. Don't panic if you don't know Java, since this section acts as a cookbook. Besides: the basic structure of the Java language is quite easy to get the hang of, especially if you have previous programming experience, whether it be in C, Pascal or Perl.

The Hello World! of auto-configuration scripts

If you have learned a programming language, you probably remember one of the most basic programs simply printing the phrase Hello World!. We don't want to print anything when someone tries to go to a page, but the following example is similar to the original Hello World program in that it's the shortest piece of code that does something useful.

The following simply connects direct to the origin server for every URL, just as it would if you had no proxy-cache configured at all.

The next example gets the browser to connect to the cache server named cache.domain.example on port 3128. If the machine is down for some reason, an error message will be returned to the user.

As you may be able to guess from the above, returning text with a semicolon (;) splits the answer returned into two sub-strings. If the first cache server is unavailable, the second will be tried. This provides you with a failover mechanism: you can attempt a local proxy server first and, if it is down, try another proxy. If all are down, a direct attempt will be made. After a short period of time, the proxy will be retried.

A third return type is included, for SOCKS proxies, and is in the same format as the HTTP type:

return "SOCKS socks.domain.example:3128";

If you have no intranet, and require no exclusions, you should use the above autoconfig file. Configuring machines with above autoconfig file allows you to add future required exclusions very easily.

Auto-config functions

Web browsers include various built-in functions to make your autoconfig coding as simple as possible. You don't have to write the code that does a string match of the hostname, since you can use a standard function call to do a match. Not all functions are covered here, since some of them are very rarely used. You can find a complete list of autoconfig functions (with examples) at the proxy autoconfig file homepage

Example autoconfig files

The main reason that autoconfig files were invented was the sheer number of possible cache setups. It's difficult (or even impossible) to represent all of the possible combinations that a autoconfig file can provide you with.

There is no config file that will work for everyone, so a couple of config files are included here, one of which should suit your setup.

Super Proxy Script

Many large ISPs will have more than one cache server. To avoid duplicating objects, these cache servers have to communicate with one another. Consider the following;

cache1 gets a request for an object. It caches the page, and stores it on disk. An hour or so later, cache2 gets a request for the same page. To find a local copy of the object, cache2 has to query the other caches. Add more and more caches, and your number of queries goes up.

If an incoming request for a specific URL only ever went to one cache, your caches would not need to communicate with one another. A client requesting the page would always connect to cache1.

Let's assume that you have 5 caches. Splitting the Internet into five pieces would split the load across the caches almost evenly. How do you split though? By destination IP address? No, since IP's like 19?.*.*.* are much more common than "5.*.*.*". By domain? No again, since one domain like would mean that you were distributing load incorrectly.

Some of you will know what a hash function is. If not, don't panic: you can still use the Super Proxy script without knowing the theoretical basis of the algorithms involved.

The Super Proxy Script allows you to split up the Internet by URL (the combination of hostname, path and filename). If you have 5 cache servers, you split up the domain of possible answers into 5 parts. (A hash function returns a number, so we are using the appropriate terms - a domain is not an Internet domain in this context). With a good hashing function, the numbers returned are going to be spread across the 5 parts evenly, which spreads your load perfectly.

If you have a cache which is twice as powerful as your others, you can allocate it more of the domain, and put more load on it.

Carp is used by some cache servers (most notably Microsoft Proxy and Squid) to decide which parent cache to send a request too. Browsers can also use CARP to decide which cache to talk to, using a java auto-config script. For more information (and an example Java script), you should look at the Super Proxy Script web page.

Future directions

There has recently been a move towards a standard for the automatic configuration of proxy-caches. New versions of Netscape and Internet Explorer are expected to use the new unknown standard to automatically change their proxy settings. This allows you to manipulate your cache server settings without inconveniencing clients.


Currently there is a major trend towards transparent caching, not only in the "Outer Internet" (where bandwidth is very expensive), but in the USA. (Transparency is covered in detail in chapter 12.)

Transparency has one major advantage: Users do not have to configure their browsers to access the cache.

To backbone providers this means that they can cache all passing traffic. A local ISP would configure their clients to talk to their cache; a backbone provider could then ask their ISP clients to use theirs as parents, but transparent caching has another advantage.

A backbone provider is acting as transit for requests that originate on other backbone provider's networks. With transparency, a backbone provider reduces this traffic as well as requests from their network to other backbone providers.

Assume you place a cache the hop before a major peering point. Here the cache intercepts both incoming requests (from other providers to web servers on your network) and outgoing (from your network to web servers on other provider's networks). This will reduce your peering-point usage (by caching outgoing requets for pages), and will also reduce the money you spend on other people's customers: since you reduce the cost it takes for data to flow out of your network. The latter cost may be minimal, but in times of network trouble it can reduce your latency noticibly.

As more and more backbone providers cache pages, more local ISPs will cache ("since it's cached further along the path, we may as well implement caching here - it's not going to change anything"). Though this will probably cause a drop in the hit rate of the backbone providers, their ever increasing user-base may make up for it. Backbone providers are caching centrally - with large numbers of edge caches (local ISP caches), they are likely to see fewer hits. Certain Inter-University networks have already noticed such a hit rate decline. As more and more universities add local caches, their hit rate falls.

Since the Universities are large, it's likely that their users will surf the same web page twice. Previously the Inter-University network would have returned the hit for that page, now the University's local cache does; this reduces the edge-cache's number of queries, and hence it's hit rate.

Chapter 6. Access Control and Access Control Operators

Access control lists (acls) are often the most difficult part of the configuration of a Squid cache: the layout and concept is not immediately obvious to most people. Hang on to your hat!

Unless chapter 3 is still fresh in your mind, you may wish to skip back and review the access control section of that chapter before you continue. This chapter assumes that you understood the difference between an acl and an acl-operator.

Access Classes and Operators

There are two elements to access control: classes and operators. Classes are defined with the acl squid.conf tag, while the names of the operators vary: the most common operator used is http_access.

Let's work through the below example line-by-line. Here, a systema administrator is in the process of installing a cache, and doesn't want other staff to access it while it's being installed, since it's likely to ping-pong up and down during the installation. Once the administrator is happy with the config, the whole network will be allowed access. The admin's PC is at the IP

If the admin connects to the cache from the PC, Squid does the following:

If you connect from a different PC (on the 10.0.*.* network) things are very similar:

If someone reaches your cache from another netblock (from, say, 192.168.*.*), the above access list will not block access. The reason for this is quite complicated. If Squid works through a set of acl-operators and finds no match, it defaults to using the opposite of the last match (if the previous operator is an allow, the default is to deny; if it's a deny, the default is to allow). This seems a bit strange at first, but let's look at an example where this behaviour is used: it's more sensible than it seems.

The following acl example is nice and simple: it's something a first-time cache admin could create.

A config file with no access lists will allow cache access without any restrictions. An administrator using the above access lists obviously wishes to allow only his network access to the cache. Given the Squid behavior of inverting the last decision, we have an invisible line reading

http_access deny all

Inverting the last decision is a simple (if not immediately obvious) solution to one of the most common acl mistakes: not adding a final deny all to the end of your acl list.

With this new knowledge, have a look at the first example in this chapter: you will see why I said not to use it in your configs. Given that the last operator denies the local network, local people will not be able to access the cache. The remainder of the Internet, however, will! As discussed in chapter 1, the simplest way of creating a catch-all acl is to match requests when they come from any IP address. When programs do netmask arithmetic a subnet of all zeros will match any IP address. A corrected version of the first example dispenses with the myNet acl.

Once the cache is considered stable and is moved into production, the config would change. http_access lines do add a very small amount of overhead, but that's not the only reason to have simple access rulesets: the less rulesets, the easier your setup is to understand. The below example includes a deny all rule although it doesn't really need one: you may know of the automatic inversion of the last rule, but someone else working on the cache may not.

You should always end your access lists with an explicit deny. In Squid-2.1 the default config file does this for you when you insert your HTTP acl operators in the appropriate place.

Acl lines

The Examples so far have given you an idea of an acl line's layout. Their layout can be symbolized as follows (? Check! ?):

acl name type (string|"filename") [string2] [string3] ["filename2"]

The acl tag consists of a minimum of three fields: a unique name; an acl type and a decision string. An acl line can have more than one decision string, hence the [string2] and [string3] in the line above.

Decision String

The acl code uses this string to check if the acl matches a given connection. When using this field, Squid checks the type field of the acl line to decide how to use the decision string. The decision string could be an IP address range, a regular expression or a list of domains or more. In the next section (where we discuss the types of acls available) we discuss the different forms of the Decision String.

If you have another look at the formal definition of the acl line above, you will note that you can have more than one decision string per acl line. Strings in this format are ORd together; if you were to specify two IP address ranges on the same line the return result of the acl would be true if either of the IP addresses match. (If source strings were ANDd together, then an incoming request would have to come from two IP address ranges at the same time. This is not impossible, but would almost certainly be pointless.)

Large decision lists can be stored in files, so that your squid.conf doesn't get cluttered. Some of the caches I have worked on have had in the region of 2000 lines of acl rules, which could lead to a very cluttered squid.conf file. You can include a file into the decision section of an acl list by placing the filename (with path) in double-quotes. The file simply contains the data set; one datum per line. In the next example the file /usr/local/squid/conf/data/myNets can contain any number of IP ranges, one range per line.

While on the topic of long lists of acls: it's important to note that you can end up slowing your cache response with very long lists of acls. Checking acls requires CPU time, and long lists can decrease cache performance, since instead of moving data to clients Squid is busy checking access lists. What constitutes a long list? Don't worry about lists with a few hundred entries unless you have a really slow or busy CPU. Lists thousands of lines long can, however, cause problems.

Types of acl

So far we have only spoken about acls that filter by source IP address. There are numerous other acl types:

Source/Destination Domain

Squid can also limit requests by their source domain. Though it doesn't always happen in the real world, network administrators can add reverse DNS entries for each of the hosts on their network. (These records are normally referred to as PTR records.) Squid can make decisions about the validity of incoming requests by checking their reverse DNS entries. In the below example, the acl is true if the request comes from a host with a reverse entry that is in either the or domains.

acl myDomain srcdomain
acl allow myDomain

Reverse DNS matches should not be used where security is important. A determined attacker (who controlled the reverse DNS entries for the attacking host) would be able to manipulate these entries so that the request comes from your domain. Squid doesn't attempt to check that reverse and forward DNS entries match, so this option is not recommended.

Squid can also be configured to deny requests to specific domains. Many people implement these filter lists for pornographic sites. The legal implications of this filtering are not covered here: there are many, and the relevant law is in a constant state of flux, so advice here would likely be obsolete in a very short period of time. I suggest that you consult a good lawyer if you want to do something like this.

The dst acl type allows one to match accesses by destination domain. This could be used to match urls for popular adult sites, and refuse access (perhaps during specific times).

If you want to deny access to a set of sites, you will need to find out these site's IP addresses, and deny access to these IP addresses too. If you just put the IP addresses in, someone determined to access a specific site could find out the IP address associated with that hostname and access it by entering the IP address in their browser.

The above is best described with an example. Here, I assume that you want to restrict access to the site www.adomain.example. If you use either the host of nslookup commands, you would find that this server has the IP address It's easiest to just have two acls: one for IPs and one for domains. If the lists get to large, you can simply place them in a file.

Words in the requested URL

Most caches can filter out URLs that contain a set of banned words. Regular expressions allow you to simply check if a word is in a given URL, but they also allow for more powerful searches of the URL. With a simple word check you would find it nearly impossible to create a rule that allows access to sites with the word sex in the URL, but at the same time denies access to all avi files on that site. With regular expressions this sort of checking becomes easy, once you understand the regex syntax.

Using Regular expressions to match words in the requested URL

Using regular expressions allows you to create more flexible access lists. So far you have only been able to filter sites by destination domain, where you have to match the entire domain to deny access to the site. Since regular expressions are used to match text strings, you can use them to match words, partial words or patterns in URLs or domains.

The most common use of regex filters in ACL lists is for the creation of far-reaching site filters: if the url or domain contain a set of banned words, access to the site is denied. If you wish to deny access to sites that contain the word sex in the URL, you would add one acl rule, rather than trying to find every site that has adult material on it.

The big problem with regex filters is that not all sites that contain the word sex in the URL are pornographic. By denying these sites you are likely to be infringing people's rights, and you should refer to a lawyer for advice on the legality of this.

Creating a list of sites that you don't want accessed can be tedious. There are companies that sell adult/unwanted material lists which plug into Squid, but these can be expensive. If you cannot justify the cost, you can

The url_regex acl type is used to match any word in the URL. Here is an example:

In places where bandwidth is very expensive, system administrators may have no problem with people visiting pornograpic sites. They may, however, want to stop people downloading huge avi files from these sites. The following example would deny downloads of avi files from sites that contain the word sex in the URL. The regular expression below matches any URL that contains the word sex AND ends with .avi.

The urlpath_regex acl strips off the url-type and hostname, checking instead only the path and filename.

Current day/time

Squid allows one to allow access to specific sites by time. Often businesses wish to filter out irrelevant sites during work hours. The Squid time acl type allws you to filter by the current day and time. By combining the dstdomain and time acls you can allow access to specific sites (such as your the sites of suppliers or other associates) during work hours, but allow access to other sites after work hours.

The layout is quite compact:

acl name time [day-list] [start_hour:minute-end_hour:minute]

Day list is a list of single characters indicating the days that the acl applies to. Using the first letter of the day would be ambiguous (since, for example, both Tuesday and Thursday start with the same letter). When the first letter is ambiguous, the second letter is used: T stands for Tuesday, H for Thursday. Here is a list of the days with their single-letter abreviations:

S - Sunday M - Monday T - Tuesday W - Wednesday H - Thursday F - Friday A - Saturday

Start_hour and end_hour are values in military time (17:00 instead of 5:00). End_hour must always be larger than start_hour; this means (unfortunately) that you cannot do the following:

# since start_time must be smaller than end_time, this won't work:
acl darkness 17:00-6:00

The only alternative to the darkness example above is something like this:

acl night time 17:00-24:00
acl early_morning time 00:00-6:00

As you can see from the original definition of the time acl, you can specify the day of the week (with no time), the time (with no day), or both the time and day (?check!?). You can, for example, create a rule that specifies weekends without specifying that the day starts at midnight and ends at the following midnight. The following acl will match on either Saturday or Sunday.

acl weekends time SA

The following example is too basic for real-world use. Unfortunately creating a good example requires some of the more advanced features of the http_access line; these are covered in the next section of this chapter, and examples are included there.

Destination Port

Web servers almost always listen for incoming requests on port 80. Some servers (notably site-specific search engines and unofficial sites) listen on other ports, such as 8080. Other services (such as IRC) also use high-numbered ports. Because of the way HTTP is designed, people can connect to things like IRC servers through your cache servers (even though the IRC protocol is very different to the HTTP protocol). The same problems can be used to tunnel telnet connections through your cache server. The major part of the HTTP specification that allows for this is the CONNECT method, which is used by clients to connect to web servers using SSL.

Since you generally don't want to proxy anything other than the standard supported protocols, you can restrict the ports that your cache is willing to connect to. The default Squid config file limits standard HTTP requests to the port ranges defined in the Safe_ports squid.conf acl. SSL CONNECT requests are even more limited, allowing connections to only ports 443 and 563.

Port ranges are limited with the port acl type. If you look in the default squid.conf, you will see lines like the following:

acl Safe_ports port 80 21 443 563 70 210 1025-65535

The format is pretty straight-forward: destination ports 443 OR 563 are matched by the first acl, 80 21 443, 563 and so forth by the second line. The most complicated section of the examples above is the end of the line: the text that reads "1024-65535".

The "-" character is used in squid to specify a range. The example thus matches any port from 1025 all the way up to 65535. These ranges are inclusive, so the second line matches ports 1025 and 65535 too.

The only low-numbered ports which Squid should need to connect to are 80 (the HTTP port), 21 (the FTP port), 70 (the Gopher port), 210 (wais) and the appropriate SSL ports. All other low-numbered ports (where common services like telnet run) do not fall into the 1024-65535 range, and are thus denied.

The following http_access line denies access to URLs that are not in the correct port ranges. You have not seen the ! http_access operator before: it inverts the decision. The line below would read "deny access if the request does not fall in the range specified by acl Safe_ports" if it were written in english. If the port matches one of those specified in the Safe_ports acl line, the next http_access line is checked. More information on the format of http_access lines is given in the next section Acl-operator lines.

http_access deny !Safe_ports


HTTP can be used for downloading (GETting data) or uploads (POSTing data to a site). The CONNECT mode is used for SSL data transfers. When a connection is made to the proxy the client specifies what kind of request (called a method) it is sending. A GET request looks like this:


If you were connecting using SSL, the GET word would be replaced with the word CONNECT.

You can control what methods are allowed through the cache using the post acl type. The most common use is to stop CONNECT type requests to non-SSL ports. The CONNECT method allows data transfer in any direction at any time: if you telnet to a badly configured proxy, and enter something like the following, you could end up connected to a machine if you had telnetted there from the cache server. This could get around packet-filters, firewall access lists and passwords, which is generally considered a bad thing!

CONNECT www.domain.example:23 HTTP/1.1

Since CONNECT requests can be quite easily exploited, the default squid.conf denies access to SSL requests to non-standard ports, as we spoke about in the previous section (on the port acl-operator.)

Let's assume that you want to stop your clients from POSTing to any sites (note that doing this is not a good idea, since people using some search engines (for example) would run into problems: at this stage this is just an example.

User name

Logs generally show the source IP address of a connection. When this address is on a multiuser machine (let's use a Unix machine at a university as an example) you cannot pin down a request as being from a specific user. There could be hundreds of people logged into the Unix machine, and they could all be using the cache server. Trying to track down a misbehaver is very difficult in this case, since you can never be sure which user is actually doing what. To solve this problem, the ident protocol was created. When the cache server accepts a connection, it can connect back to the origin server (on a low-numbered port, so the reply cannot be faked) and finds out who just connected. This doesn't make any sense on windows systems: people can just load their own ident servers (and become daffy duck for a day). If you run multi-user systems then you may want only certain people on those machines to be able to use the cache. In this case you can use the ident username to allow or deny access.

One of the best things about Unix is the flexibility you get. If you wanted (for example) only students in their second year on to have access to the cache servers via your Unix machines, you could create a replacement ident server. This server could find out which user that has connected to the cache, but instead of returning the username you could return a string like "third_year" or "postgrad". Rather than maintaining a list of which students are in on both the cache server and the central Unix system, you could simple Squid rules, and the ident server could do all the work where it checks which user is which.

Username/Password pair

If you want to track Internet usage it's best to get users to log into the cache server when they want to use the net. You can then use a stats program to generate per-user reports, no matter which machine on your network a person is using. Universities and colleges often have labs with many machines, where it is difficult to tell which user is sitting in front of a machine at any specific time. By using names and passwords you will solve this problem.

Squid uses modules to do user authentication, rather than including code to do it directly. The default Squid source does, however, include two standard modules; The first authenticates users from a file, the other uses SMB (Windows NT) authentication. These modules are in the auth_modules directory in the source directory. These modules are not compiled when you compile Squid itself, and you will need to chooes an authentication module and run make in the appropriate directory. If the compile goes well, a make install will place the program file in the /usr/local/squid/bin/ directory and any config files in the /usr/local/squid/etc/ directory.

NCSA authentication is the easiest to use, since it's self contained. The SMB authentication program requires that SAMBA be installed, since it effectively talks to the NT server through SAMBA.

The squid.conf file uses the authenticate_program tag to decide which external program to use to authenticate users. If Squid were to only start one authentication program, a slow username/password lookup could slow the whole cache down (while all other connections waited to be authenticated). Squid thus opens more than one authentication program at a time, sending pending requests to the second when the first is busy, the third when the second is and so forth. The actual number started is specified by the authenticate_children squid.conf value. The default number started is five, but if you have a heavily loaded cache then you will need to increase this value.

Acl-operator lines

Acl-operators are the other half of the acl system. For each connection the appropriate acl-operators are checked (in the order that they appear in the file). You have met the http_access and icp_access operators before, but they aren't the only Squid acl-operators. All acl-operator lines have the same format; although the below format mentions http_access specifically, the layout also applies to all the other acl-operators too.

http_access allow|deny [!]aclname [& [!]aclname2 ... ]

Let's work through the fields from left to right. The first word is http_access, the actual acl-operator.

The allow and deny words come next. If you want to deny access to a specific class of users, you can change the customary allow to deny in the acl line. We have seen where a deny line is useful before, with the final deny of all IP ranges in previous examples.

Let's say that you wanted to deny Internet access to a specific list of IP addresses during the day. Since acls can only have one type per acl, you could not create an acl line that matches an IP address during specific times. By combining more than one acl per acl-operator line, though, you get the same effect. Consider the following acls:

acl dialup src
acl work time 08:00-17:00

If you could create an acl-operator that was matched when both the dialup and work acls were true, clients in the range could only connect during the right times. This is where the aclname2 in the above acl-operator definition comes in. When you specify more than one acl per acl-operator line, both acls have to be matched for the acl-operator to be true. The acl-operator function AND's the results from each acl check together to see if it is to return true of false.

You could thus deny the dialup range cache access during working hours with the following acl rules:

You can also invert an acl's result value by using an exclamation mark (the traditional NOT value from many programming languages) before the appropriate acl. In the following example I have reduced Example 6-4 into one http_access line, taking advantage of the implicit inversion of the last rule to deny access to all clients.

Since the above example is quite complicated: let's cover it in more detail:

In the above example an IP from the outside world will match the 'all' acl, but not the 'myNet' acl; the IP will thus match the http_access line. Consider the binary logic for a request coming in from the outside world, where the IP is not defined in the myNet acl.

Deny http access if ((true) & (!false))

If you consider the relevant matching of an IP in the range, the myNet value is true, the binary representation is as follows:

Deny http access if ((true) & (!true))

A range IP will thus not match the only http_access line in the squid config file. Remembering that Squid will default to using the inverse of the last match in the file, accesses will be allowed from the myNet IP range.

The other Acl-operators

You have encountered only the http_access and icp_access acl-operators so far. Other acl-operators are:

The ident_lookup_access acl-operator

Earlier we discussed using the ident protocol to control cache access. To reduce network overhead, Squid does an ident lookup only when it needs to. If you are using ident to do access control, Squid will do an ident lookup for every request, and you don't have to worry about this acl-operator.

Many administrators would like to log the the ident value for connections without actually using it for access control. Squid used to have a simple on/off switch for ident lookups, but this incurred extra overhead for the cases where the ident lookup wasn't useful (where, for example, the connection is from a desktop PC).

Let's consider some examples. Assume that a you have one Unix server (at IP address, and all remaining IP's in the range are desktop PC's. You don't want to log the ident value from PC's, but you do want to record it when the connection is from the Unix machine. Here is an example acl set that does this:

If a system cracker is attempting to attack your cache, it can be useful to have their ident value logged. The following example gets Squid not to do ident lookups for machines that are allowed access, but if a request comes from a disallowed IP range, an ident lookup is done and inserted into the log.

The miss_access acl-operator

The ICP protocol is used by many caches to find out if objects are in another cache's on-disk store. If you are peering with other organisation's caches, you may wish them to treat you as a sibling, where they only get data that you already have stored on disk. If an unscrupulous cache-admin were to change their cache_peer line to read parent instead of sibling, they could get you to retrieve objects on their behalf.

To stop this from happening, you can create an acl that contains the peering caches, and use the miss_access acl-operator to ensure that only hits are served to these caches. In response to all other requests, an access-denied message is sent (so if a sibling complains that they almost always get error messages, it's likely that they think that you should be their parent, and you think that they should be treating you as a sibling.)

When looking at the following example it is important to realise that http_access lines are checked before any miss_access lines. If the request is denied by the http_access lines, an error page is returned and the connection closed, so miss_access lines are never checked. This means that the last miss_access line in the example doesn't allow random IP ranges to access your cache, it only allows ranges that have passed the http_access test through. This is simpler than having one miss_access line for each http_access line in the file, and it will reduce CPU usage too, since only two acls are checked instead of the six we would have instead.

SNMP Configuration

Before we continue: if you wish to use Squid's SNMP functions, you will need to have configured Squid with the --enable-snmp option, as discussed way back in Chapter 2. The Squid source only includes SNMP code if it is compiled with the correct options.

Normally a Unix SNMP server (also called an agent) collects data from the various services running on a machine, returning information about the number of users logged in, the number of sendmail processes running and so forth. As of this writing, there is no SNMP server which gathers Squid statistics and makes them available to SNMP managment stations for interpretation. Code has thus been added to Squid to handle SNMP queries directly.

Squid normally listens for incoming SNMP requests on port 3401. The standard SNMP port is 161.

For the moment I am going to assume that your management station can collect SNMP data from a port other than 161. Squid will thus listen on port 3401, where it will not interfere with any other SNMP agents running on the machine.

No specific SNMP agent or mangement station software is covered by this text. A Squid-specific mib.txt file is included in the /usr/local/squid/etc/ directory. Most management station software should be able to use this file to construct Squid-specific queries.

Running multiple SNMP servers on a cache machine

If you are running multiple SNMP servers on your cache machine, you probably want to see all the SNMP data returned on one set of graphs or summaries. You don't want to have to query two SNMP servers on the same machine, since many SNMP analysis tools will not allow you to relate (for example) load average to number of requests per second when the SNMP data comes from more than one source.

Let's work through the steps Squid goes through when it receives an SNMP query: The request is accepted, and access-control lists are checked. If the request is allowed, Squid checks to see if it's a request for Squid information or a request for something it doesn't understand. Squid handles all Squid-specific queries internally, but all other SNMP requests are simply passed to the other SNMP server; Squid essentially acts as an SNMP proxy for SNMP queries it doesn't understand.

This SNMP proxy-mode allows you to run two servers on a machine, but query them both on the same port. In this mode Squid will normally listen on port 161, and the other SNMP server is configured to listen on another port (let's use port 3456 for argument's sake). This way the client software doesn't have to be configured to query a different port, which especially helps when the client is not under your control.

Binding the SNMP server to a non-standard port

Getting your SNMP server to listen on a different port may be as easy as changing one line in a config file. In the worst case, though, you may have to trick it to listen somewhere else. This section is a bit of a guide to IP server trickery!

Server software can either listen for connections on a hard-coded port (where the port to listen to is coded into the source and placed directly into the binary on compilation time), or it can use standard system calls to find the port that it should be listening to. Changing programs that use the second set of options to use a different port is easy: you edit the /etc/services file, changing the value for the appropriate port there. If this doesn't work, it probably means that your program uses hard-coded values, and your only recourse is to recompile from source (if you have it) or speak to your vendor.

You can check that your server is listening to the new port by checking the output of the netstat command. The following command should show you if some process is listening for UDP data on port 3456:

cache1:~ $ netstat -na | grep udp | grep 3456
udp        0      0   *
cache1:~ $ 

Changing the services port does have implications: client programs (like any SNMP management station software running on the machine) will also use the services file to find out which port they should connect when forming outgoing requests. If you are running anything other than a simple SNMP agent on the cache machine, you must not change the /etc/services file: if you do you will encounter all sorts of strange problems!

Squid doesn't use the /etc/services file, but the port to listen to is stored in the standard Squid config file. Once the other server is listening on port 3456, we need to get Squid to listen on the standard SNMP port and proxy requests to port 3456.

First, change the snmp_port value in squid.conf to 161. Since we are forwarding requests to another SNMP server, we also need to set forward_snmpd_port to our other-server port, port 3456.

Access Control with more than one Agent

Since Squid is actually creating all the queries that reach the second SNMP server, using an IP-based access control system in the second server's config is useless: all requests will come from localhost. Since the second server cannot find out where the requests came from originally, Squid will have to take over the access control functions that were handled by the other server.

For the first example, let's assume that you have a single SNMP management station, and you want this machine to have access to all SNMP functions. Here we assume that the management station is at IP

You may have classes of SNMP stations too: you may wish some machines to be able to inspect public data, but others are to be considered completely trusted. The special snmp_community acl type is used to filter requests by destination community. In the following example all local machines are able to get data in the public SNMP community, but only the snmpManager machine is able to get other information. In this example we are using the ANDing of the publicCommunity and myNet acls to ensure that only people on the local network can get even public information.

Delay Classes

Delay Classes are generally used in places where bandwidth is expensive. They let you slow down access to specific sites (so that other downloads can happen at a reasonable rate), and they allow you to stop a small number of users from using all your bandwidth (at the expense of those just trying to use the Internet for work).

Many non-US Universities have very small pipes to the Internet. Unfortunately these Universities often end up with huge amounts of their bandwidth being used for surfing that is not study-related. In the US this is fine, since the cost is negligible, but in other countries the cost of this casual surfing is astronomical.

To ensure that some bandwidth is available for work-related downloads, you can use delay-pools. By classifying downloads into segments, and then allocating these segments a certain amount of bandwidth (in kilobytes per second), your link can remain uncongested for useful traffic.

To use delay-pools you need to have compiled Squid with the appropriate source code: you will have to have used the --enable-delay-pools option when running the configure program back in Chapter 2.

The Second Pool Class

Rather than cover all of the available classes immediately, let's deal with a basic example first. In this example we have only one pool, and the pool catches all URLs containing the word abracadabra.

The first line is a standard ACL: it returns true if the requested URL has the word abracadabra in it. The -i flag is used to make the search case-insensitive.

The delay_pool_count variable tells Squid how many delay pools there will be. Here we have only one pool, so this option is set to 1.

The third line creates a delay pool (delay pool number 1, the first option) of class 1 (the second option to delay_class).

The first delay class is the simplest: the download rate of all connections in the class are added together, and Squid keeps this aggregate value below a given maximum value.

The fourth line is the most complex, as if you can see. The delay_parameters option allows you to set speed limits on each pool. The first option is the pool to be manipulated: since we have only one pool in this example, this is set to 1. The second option consists of two values: the restore and max values, seperated by a forward-slash (/).

If you download a short file at high speed, you create a so-called burst of traffic. Generally these short bursts of traffic are not a problem: these are normally html or text files, which are not the real bandwidth consumers. Since we don't want to slow everyone's access down (just the people downloading comparitively large files), Squid allows you to configure a size that the download is to start slowing down at. If you download a short file, it arrives at full speed, but when you hit a certain threshold the file arrives more slowly.

The restore value is used to set the download speed, and the max value lets you set the size at which the files are to be slowed down from. Restore is in bytes per second, max is in bytes.

In the above example, downloads proceed at full speed until they have downloaded 16000 bytes. This limit ensures that small file arrive reasonably fast. Once this much data has been transferred, however, the transfer rate is slowed to 16000 bytes per second. At 8 bits per byte this means that connections are limited to 128kilobits per second (16000 * 8).

The Second Pool Class

As I discussed in this section's introduction, delay pools can help you stop one user from flooding your links with downloads. You could place each user in their own pool, and then set limits on a per-user basis, but administrating these lists would become painful almost immediately. By using a different pool type, you can set rate limits by IP address easily.

Let's consider another example: you have a 128kbit per second line. Since you want some bandwidth available for things like SMTP, you want to limit web access to 100kbit per second. At the same time, you don't want a single user to use more than their fair share of sustained bandwidth. Given that you have 20 staff members, and 100kbit per second remaining bandwidth, each person should not use more than 5kbit per second of bandwidth. Since it's unlikely that every user will be surfing at once, we can probably limit people to about four times their limit (that's 20kbit per second, or 2.5kbytes per second).

In the following example, we change the delay class for pool 1 to 2. Delay class 2 allows us to specify both an aggregate (overall) bandwidth usage and a per-user usage. In the previous example the delay_paramaters tag only took one set of options, the aggregate peak and burst rates. Given that we are now using a class-two pool, we have to supply two sets of options to delay_parameters: the overall speed and the per-IP speed. The 100kbits per second value is converted to bytes per second by dividing by 8 (giving us the 12500 values), and the per-IP value of 2.5kbits per second we discovered is converted to bytes per second (giving us the 2500 values.)

The Third Pool Class

This class is useful to very organizations like Universities. The second pool class lets you stop individual users from flooding your links. A lab full of students all operating at their maximum download rate can, however, still flood the link. Since such a lab (or department, if you are not at a University) will all have IP addresses in the same range, it is useful to be able to put a cap on the download rate of an entire network range. The third pool class lets you do this. Currently this option only works on class-C network ranges, so if you are using variable length subnet masks then this will not help.

In the next example we assume that you have three IP ranges. Each range must not use more than 1/3 of your available bandwidth. For this example I am assuming that you have a 512kbit/s line, and you want 64kbit/s available for SMTP and other protocols. This will leave you with an overall download rate cap of 448kbit/s.) Each Class-C IP range will have about 150kbit/s available. With 3 ranges of 256 IP addresses each, you should have in the region of 500 pc's, which (if calculated exactly) gives you .669kbit per second per machine. Since it is unlikely that all machines will be using the net at the same time, you can probably allocate each machine (say) 4kbit per second (a mere 500 bytes per second).

In this example, we changed the delay class of the pool to 3. The delay_parameters option now takes four arguments: the pool number; the overall bandwidth rate; the per-network bandwidth rate and the per-user bandwidth rate.

The 4kbit per second limit for users seems a little low. You can increase the per-user limit, but you may find that it's a better idea to change the max value instead, so that the limit sets in after only (say) 16kilobytes or so. This will allow small pages to be downloaded as fast as possible, but large pages will be brought down without influencing other users.

If you want, you can set the per-user limit to something quite high, or even set them to -1, which effectively means that there is no limit. Limits work from right to left, so if I user is sitting alone in a lab they will be limited by their per-user speed. If this value is undefined, they are limited by their per-network speed, and if that is undefined then they are limited by their overall speed. This means that you can set the per-user limit higher than you would expect: if the lab is not busy then they will get good download rates (since they are only limited by the per-network limit).

Chapter 7. Cache Hierarchies

Why Peer

The primary function of an inter-cache protocol is to stop object duplication, increasing hit rates. If you have a large network with widely separated caches, you may wish to store objects in each cache even if one of your other caches has it: by keeping objects close to your users, you reduce their network latency (even if you end up "wasting" disk space in the process.)

Inter-branch traffic can be reduced by placing a cache at each branch. Since caches can avoid duplicating objects between them, each disk you add to a cache adds space to the overall hierarchy, increasing your hierarchy hit-rate. This is a lot better than simply having caches at branches which do not communicate with one another, since with that setup you end up end up with multiple copies of each cache object; one per server. Clients can also be configured to query another branches cache if their local one goes down, adding redundancy.

If overloaded, a central cache machine can become a network bottleneck. Unlike one cache machine, caches in a hierarchy can be close to all parts of the network; they can also handle a much larger load (with a near-linear increase in performance with each added machine). Loaded caches can thus be replaced with clusters of low-load caches, without wasting disk space.

Integrating your caches into a public cache hierarchy can increase your hit rate (since you increase your effective disk space by accessing other machine's object stores.) By choosing peers carefully, you can reduce latency, or reduce costs by saving Internet bandwidth (if communicating with your peers is cheaper than going direct to the source.) On the other hand, communicating with peers via loaded (or high-latency) line can slow down your cache. It's best to check your peer response times periodically to check if the peering arrangement is beneficial. You can use the client program to check cache response times, and the cache manager (discussed in Chapter 12) to look at Squid's view on the cache.

Peer Configuration

First, let's look at the squid.conf options available for hierarchy configuration. We will then work through the most common hierarchy structures, so that you can see the way that the options are used.

You use the cache_peer option to configure the peers that Squid will communicate with. Other options are then used to select which peer to pass a request to.

The cache_peer Option

When communicating with a peer, Squid needs some basic information about how to talk to the machine; the hostname, what ports to send queries to, and so forth. The cache_peer config line does this. Let's look at an example line:

The cache_peer option is split into five fields. The first field (cache.domain.example) is the hostname or IP of the cache that is to be queried. The second field indicates the type of relationship, and must be set to either parent or sibling or multicast. The third field sets the HTTP port of the destination server, while the fourth sets the ICP (UDP) query port. The fifth field can contain more than zero or more keywords, although we only use one in the example above; the keyword default sets that the cache will be used as the default path to the outside world. If you compiled Squid to support HTCP, your cache will automatically attempt to connect to TCP port 4827 (there is currently no option to change this port value). Cache digests are transferred via the HTTP port specified on the cache_peer line.

Here is a summary of the available cache_peer options:

Peer Selection

Let's say that you have only one parent cache server: the server at your ISP. In Chapter 3, we configured Squid so that the parent cache server would not be queried for internal hosts, so queries to the internal machines went direct, instead of adding needless load to your parent cache (and the line between you). Squid can use access-control lists to decide which cache to talk to, rather than just the destination domain. With access lists, you can use different caches depending on the source IP, domain, text in the URL and more. The advantages of this flexibility are not immediately obvious (even to me), but some examples are given in th remainder of this chapter. First, however, let's cover filtering by destination domain.

Selecting with Acls

Squid can also make peer selections based on the results of acl rules. The cache_peer_access line is discussed in the previous chapter. The following example could be used if you want all requests from a specific IP address range to go to a specific cache server (for accounting purposes, for example). In the following example, all requests from the 10.0.1.* range are passed to cache.domain.example, but all other requests are handled directly.

The always_direct and never_direct tags

Squid checks all always_direct tags before it checks any never_direct tags. If a matching always_direct tag is found, Squid will not check the never_direct tags, but decides which cache to talk to immediately. This behavior is demonstrated by the following example; here, Squid will attempt to go the machine intranet, even though the same host is also matched by the all acl.

Let's work through the logic that Squid uses in the above example, so that you can work out which cache Squid is going to talk to when you construct your own rules.

First, let's consider a request destined for the web server intranet.mydomain.example. Squid first works through all the always_direct lines; the request is matched by the first (and only) line. The never_direct and always_direct tags are acl-operators, which means that the first match is considered. In this illustration, the matching line instructs Squid to go directly when the acl matches, so all neighboring peers are ignored for this request. If the line used the deny keyword instead of allow, Squid would have simply skipped on to checking the never_direct lines.

Now, the second case: a request arrives for an external host. Squid works through the always_direct lines, and finds that none of them match. The never_direct lines are then checked. The all acl matches the connection, so Squid marks the connection as never to be forwarded directly to the origin server. Squid then works through it's list of peers, trying to find the cache that the request is best forwarded to (servers that have the object are more likely to get the request, as are servers that respond fast). The algorithm that Squid uses to decide which of it's peers to use is discussed shortly.

Multicast Cache Communication

Cache digests are in some ways a replacement for multicast cache peering. There are some advantages to cache-digests: they are handled at the Squid level (so you don't have to fiddle with kernel multicast settings and so forth), and they add significantly less latency (finding out if a cache has an object simply involves checking an in-memory bit-array, which is significantly faster than checking across the network).

First, though, let's cover some terminology. Most people are familiar with the term broadcast, where data is sent from one host to all hosts on the local network. Broadcasts are normally used to discover things, not for general inter-machine transfer: a machine will send out a broadcast ARP request to try and find the hardware address that a specific IP address belongs to. You can also send ping packets to the broadcast address, and find machines on the local network when they respond. Broadcasts only work across physical segments (or bridged/switched networks), so an ARP request doesn't go to every machine on the Internet.

A unicast packet is the complete opposite: one machine is talking to only one other machine. All TCP connections are unicast, since they can only have one destination host for each source host. UDP packets are almost always unicast too, though they can be sent to the broadcast address so that they reach every single machine in some cases.

A multicast packet is from one machine to one or more. The difference between a multicast packet and a broadcast packet is that hosts receiving multicast packets can be on different lans, and that each multicast data-stream is only transmitted between networks once, not once per machine on the remote network. Rather than each machine connecting to a video server, the multicast data is streamed per-network, and multiple machines just listen-in on the multicast data once it's on the network.

This efficient use of bandwidth is perfect for large groups of caches. If you have more than one server (for load-distribution, say), and someone wants to peer with you, they will have to configure their server to send one ICP packet to each of your caches. If Squid gets an ICP request from somewhere, it doesn't check with all of it's peers to see if they have the object. This "check with my peers" behavior only happens when an HTTP request arrives. If you have 5 caches, anyone who wants to find out if your hierarchy has an object will have to send 5 ICP requests (or treat you as a parent, so that your caches check with one another). This is a real waste of bandwidth. With a multicast network, though, the remote cache would only send one ICP request, destined for your multicast address. Routers between you would only transfer one packet (instead of 5), saving the duplication of requests. Once on your network, each machine would pick up one packet, and reply with their answer.

Multicast packets are also useful on local networks, if you have the right network cards. If you have a large group of caches on the same network, you can end up with a lot of local traffic. Each request that a cache receives prompts one ICP request to all the other local caches, swamping the local network with small packets (and their replies). A multicast packet, on the other hand, is a kind of broadcast to the machines on the local network. They will each receive a copy of the packet, although only one went out onto the wire. If you have a good ethernet card, the card will handle a fair amount of the filtering (some cards may have to be put into promiscuous mode to pick up all the packets, which can cause load on the machine: make sure that the card you buy supports hardware multicast filters). This solution is still not linearly scalable, however, since the reply packets can easily become the bottleneck by themselves.

Getting your machine ready for Multicast

The kernel's IP stack (the piece of kernel code that handles IP networking) needs to look out for multicast packets, otherwise they will be discarded (either by the network card or the lower levels of the IP stack.) Your kernel may already have multicast support, or you will have to turn it on. Doing this is, unfortunately, beyond the scope of this book, and you may have to root around for a howto guide somewhere.

Once your machine is setup to receive multicast packets, you need your machines to talk to one another. You can either join the mbone (a virtual multicast backbone), or set up an internal multicast network. Joining the mbone could be a good thing anyway, since you get access to other services. You must be sure not to use a random set of multicast IP addresses, since they may belong to someone else. You can get your own IP range from the people at the mbone.

An outgoing multicast packet has a ttl (Time To Live) value, which is used to ensure that loops are not created. Each time a packet passes through a router, the router decrements this ttl value, and the value is then checked. Once the value reaches zero, the packet is dropped. If you want multicast packets to stay on your local network, you would set the ttl value to 1. The first router to see the packet would decrement the packet, discover the ttl was zero and discard it. This value gives you a level of control on how many multicast routers will see the packet. You should set this value carefully, so that you limit packets to your local network or immediate multicast peers (larger multicast groups are seldom of any use: they generate too many responses, and when geographically dispersed, may simply add latency. You also don't want crackers picking up all your ICP requests by joining the appropriate multicast group.)

Various multicast debugging tools are available. One of the most useful is mtrace, which is effectively a traceroute program for multicast connections. This program should help you choose the right ttl value.

Cache Digests

Cache digests are one of the latest peering developments. Currently they are only supported by Squid, and they have to be turned on at compile time.

Squid keeps it's "list" of objects in an in-memory hash. The hash table (which is based on MD5) helps Squid find out if an object is in the cache without using huge amounts of memory or reading files on disk. Periodically Squid takes this table of objects and summarizes it into a small bitmap (suitable for transfer across a modem). If a bit in the map is on, it means that the object is in the store, if it's off, the object is not. This bitmap/summary is available to other caches, which connect on the HTTP port and request a special URL. If the client cache (the one that just collected the bitmap) wants to know if the server has an object, it simply performs the same mathematical function that generated the values in the bitmap. If the server has the object, the appropriate bit in the bitmap will be defined.

There are various advantages to this idea: if you have a set of loaded caches, you will find that inter-cache communication can use significant amounts of bandwidth. Each request to one cache sparks off a series of requests to all neighboring caches. Each of these queries also causes some server load: the networking stack has to deal with these extra packets, for one thing. With cache-digests, however, load is reduced. The cache digest is generated only once every 10 minutes (the exact value is tunable). The transfer of the digest thus happens fairly seldom, even if the bitmap is rather large (a few 100kbytes is common.) If you were to run 10 caches on the same physical network, however, with each ICP request being a few hundred bytes, the numbers add up. This network load reduction can give your cache time to breathe too, since the kernel will not have to deal with as many small packets.

ICP packets are incredibly simple: they essentially contain only the requested URL. Today, however, a lot of data is transferred in the headers of a request. The contents of a static URL may differ depending on the browser that a user uses, cookie values and more. Since the ICP packet only contains the URL, Squid can only check the URL to see if it has the object, not both the headers and the URL. This can (very occasionally) cause strange problems, with the wrong pages being served. With cache digests, however, the bitmap value depends on both the headers AND the url, which stops these strange hits of objects that are actually generated on-the-fly (normally these pages contain cgi-bin in their path, but some don't, and cause problems.)

Cache digests can generate a small percentage of false hits: since the list of objects is updated only every 10 minutes, your cache could expire an object a second after you download the summarized index. For the next ten minutes, the client cache would believe your server has data that it doesn't. Some five percent of hits may be false, but they are simply retrieved directly from the origin server if this turns out to be the case.

Cache Hierarchy Structures

Deciding what hierarchy structure to use is difficult. Not only that, but it's quite often very difficult to change, since you can have thousands of clients accessing you directly, and even more through your peers.

Here, I cover the most common cache-peer architectures. We start off with the most simple setup that's considered peering: two caches talking to one another as siblings. I am going to try and cover all the issues with this simple setup, and then move to larger cache meshes.

Two Peering Caches

We have two caches, your cache and their cache. You have a nice fast link between them, with low-latency, both caches have quite a lot of disk space, and they aren't going to be running into problems with load anytime soon. Let's look at the peering options you have:


The traditional cache hierarchy structure involves lots of small servers (with their own disk space, each holding the most common objects) which query another set of large parent servers (there can even be only one large server.) These large servers then query the outside on the client cache's behalf. The large servers keep a copy of the object so that other internal caches requesting the page get it from them. Generally, the little servers have a small amount of disk space, and are connected to the large servers by quite small lines.

This structure generally works well, as long as you can stop the top-level servers from becoming overloaded. If these machines have problems, all performance will suffer.

Client caches generally do not talk to one another at all. The parent cache server should have any object that the lower-down cache may have (since it fetched the object on behalf of the lower-down cache). It's invariably faster to communicate with the head-office (where the core servers would be situated) than another region (where another sibling cache is kept).

In this case, the smaller servers may as well treat the core servers as default parents, even using the no-query option, to reduce cache latency. If the head-office is unreachable it's quite likely that things may be unusable altogether (if, on the other hand, your regional offices have their own Internet lines, you can configure the cache as a normal parent: this way Squid will detect that the core servers are down, and try to go direct. If you each have your own Internet link, though, there may not be a reason to use a tree structure. You might want to look at the mesh section instead, which follows shortly.)

To avoid overloading one server, you can use the round-robin option on the cache_peer lines for each core server. This way, the load on each machine should be spread evenly.

The Cache Array Routing Protocol (CARP)

The CARP protocol uses a hash function to decide which cache a request is to be forwarded to. When a request is to be sent out, the code takes the URL requested and feeds it through a formula that essentially generates a large number from the text of the URL. A different URL (even if it differs by only one character) is likely to end up as a very different number (it won't, for example, differ by one). If you take 50 URLs and put them through the function, the numbers generated are going to be spread far apart from one another, and would be spread evenly across a graph. The numbers generated, however, all fit in a certain range. Because the number are spread across the range evenly, we can split the range into two, and the same number of URLs will have ended up in the first half as the second.

Let's say that we create a rule that effectively says: "I have two caches. Whenever I receive a request, I want to pass it to one of these caches. I know that any number generated by the hash function will be less than X, and that numbers are as likely to fall above one-half X as below. By sending all requests that hash to less than one-half X to cache one, and the remaining requests to cache two, the load should be even."

To terminology: the total range of numbers is split into equally large number ranges (called buckets).

Let's say that we have two caches, again. This time, though, cache one is able to handle twice the load of cache two. If we split the hash space into three ranges, and allocate buckets one and three to cache one (and bucket two to cache two), a URL will have twice the chance of going to cache one as it does to cache two.

Squid caches can talk to parent caches using CARP balancing if CARP was enabled when the source was configured (using the ./configure --enable-carp command.)

The load-factor values on all cache_peer lines must add up to 1.0. The below example splits 70% of the load onto the machine bigcache.mydomain.example, leaving the other 30% up to the other cache.

Now that your cache is integrated into a hierarchy (or is a hierarchy!), we can move to the next section. Accelerator mode allows your cache to function as a front-end for a real web server, speeding up web page access on those old servers.

Transparent caches effectively accelerate web servers from a distance (the code, at least, to perform both functions is effectively the same.) If you are going to do transparent proxying, I suggest that you read the next two Chapters. If you aren't interested in either of these Squid features, your Squid installation should be up and running. The remainder of the book (Section III) covers cache maintenance and debugging.

Chapter 8. Accelerator Mode

Some cache servers can act as web servers (or vis versa). These servers accept requests in both the standard web-request format (where only the path and filename are given), and in the proxy-specific format (where the entire URL is given).

The Squid designers have decided not to let Squid be configured in this way. This avoids various complicated issues, and reduces code complexity, making Squid more reliable. All in all, Squid is a web cache, not a web server.

By adding a translation layer into Squid, we can accept (and understand) web requests, since the format is essentially the same. The additional layer can re-write incoming web requests, changing the destination server and port. This re-written request is then treated as a normal request: the remote server is contacted, the data requested and the results cached. This lets you get Squid to pretend to be a web server, re-writing requests so that they are passed on to some other web server.

When to use Accelerator Mode

Accelerator mode should not be enabled unless you need it. There are a limited set of circumstances in which it is needed, so if one of the following setups applies to you, you should have a look at the remainder of this chapter.

Accelerator Configuration Options

The list of accelerator options is short, and setup is fairly simple. Once we have a working accelerator cache, you will have to create the appropriate access-list rules. (Since you probably want people outside your local network to be able to access your server, you cannot simple use source-IP address rulesets anymore.)

The httpd_accel_uses_host_header option

A normal HTTP request consists of three values: the type of transfer (normally a GET, which is used for downloads); the path and filename to be retrieved (or executed, in the case of a cgi program); and the HTTP version.

This layout is fine if you only have one web site on a machine. On systems where you have more than one site, though, it makes life difficult: the request does not contain enough information, since it doesn't include information about the destination domain. Most operating systems allow you to have IP aliases, where you have more than one IP address per network card. By allocating one IP per hosted site, you could run one web server per IP address. Once the programs were made more efficient, one running program could act as a server for many sites: the only requirement was that you had one IP address per domain. Server programs would find out which of the IP addresses clients were connected to, and would serve data from different directories for each IP.

There are a limited number of IP addresses, and they are fast running out. Some systems also have a limited number of IP aliases, which means that you cannot host more than a (fairly arbitrary) number of web sites on machine. If the client were to pass the destination host name along with the path and filename, the web server could listen to only one IP address, and would find the right destination directores by looking in a simple hostname table.

From version 1.1 on, the HTTP standard supports a special Host header, which is passed along with every outgoing request. This header also makes transparent caching and acceleration easier: by pulling the host value out of the headers, Squid can translate a standard HTTP request to a cache-specific HTTP request, which can then be handled by the standard Squid code. Turning on the httpd_accel_uses_host_header option enables this translation. You will need to use this option when doing transparent caching.

It's important to note that acls are checked before this translation. You must combine this option with strict source-address checks, so you cannot use this option to accelerate multiple backend servers (this is certain to change in a later version of Squid).

Related Configuration Options

So far, we have only covered the Config options that directly relate to accelerator mode.

Access Control

Presumably you will want people from outside your network to be able to access the web server that Squid is accelerating. If you have based your access lists on the examples in this book, you will find that machines on the outside cannot access the site being accelerated. The accelerated request is treated exactly like a normal http request, so people accesing the site from the outside world will be rejected since your acl rules deny access from IPs that are not on your network. By using the dst acl type, you can add specific exclusions to your access lists to allow requests to the accelerated host.

In the following example, we have changed the config so that the first rule matches (and allows) any request to the machine at IP, the accelerated machine. If we did not have the port acl in the below rules, someone could request a URL with a different port number with a request that explicitly specifies a non-standard port. If we were to leave out this rule, it could let a system cracker poke around the system with requests for things like http://server.mydomain.example:25.

Example Configurations

Let's cover two example setups: one, where you are simply using Squid's accelerator function so that the machine has both a web server and a cache server on port 80; two, where you are using Squid as an accelerator to speed up a slow machine.

Replacing a Combination Web/Cache server

First, let's cover the most common use of accelerator mode: replacing a combination web/cache server with Squid. When Squid is acting as an accelerator (speeding up a slow web server), Squid will accept requests on port 80 (on any IP address) and pass them to a cache server on a different machine (also on port 80). Since it's unlikely that you want to use two machines where you can use one (unless you are changing to Squid due to server overload), we will need to configure Squid to pass requests to the local machine.

Squid will need to accept incoming requests on port 80 (using the http_port option), and pass the requests on to the web server on another port (since only one process can listen for requests on port 80 at a time). I normally get web servers to listen for requests on port 8000.

Since you want Squid to function both as an accelerator and as a cache server, you will need to use the httpd_accel_with_proxy option.

The cache in this example is the local machine: there is almost certainly no reason to cache results from this server. I could have used an extremely conservative refresh_pattern in the below example, but instead I decide to use the no_cache tag: this way I can make use of my predefined acl. The always_direct tag in the below example will be very useful if you have a peer cache: you don't want the request passed on to a peer machine.

Chapter 9. Transparent Caching

When you implement disk caching in an Operating System Kernel, all applications automatically see the benefit: the data caching happens without their knowledge. Since the Operating System ensures that on-disk copies of data are always the same as the cached copies, the data that an application reads is never out of date.

With web caching, however, there is a chance that the original data can change without the cache knowing. Squid uses refresh patterns (described in chapter 11) to decide when cached objects are to be removed. If these rules are too agressive, you could end up serving stale objects to clients. Even if these rules are perfect, an incorrectly configured source-server could get Squid to return old objects. Because users could retrieve an out of date page, you should not implement caching without their knowledge.

Squid can be configured to act transparently. In this mode, clients will not configure their browsers to access the cache, but Squid will transparently pick up the appropriate packets and cache requests. This solves the biggest problem with caching: getting users to use the cache server. Users hardly ever know how to configure their browsers to use a cache, which means that support staff have to spend time with every user getting them to change their settings. Some users are worried about their privacy, or they think (that since it's a host between them and the Internet) that the cache is slower (certainly not the case, as a few tests with the client program will show).

However: transparent caching isn't really transparent. The cache setup is transparent, but using the cache isn't. Users will notice a difference in error messages, and even the progress bars that browsers show can act differently.

The Problem with Transparency

When Squid transparently caches a site, the source IP address of the connection changes: the request comes from the cache server rather than the client machine. This can play havoc with web sites that use IP-address authentication (such sites only allow requests from a small set of IP addresses, rather than authenticating requests with a name and password.)

Since the cache changes the source IP address of the connection, some servers may deny legitimate users access. In many cases, this will cost users money (they may pay for the service, or use the information on that site to make money.)

If you know your network inside out, and know exactly who would be accessing a site like this, there is probably no problem with using transparent caching. If this is the case, though, it might be easier to simply change all of your users' settings.

Dialup ISPs generally have little problem implementing transparent caching, since dialup customers almost always get a different IP address whenever they connect. They cannot thus access sites which require a static IP address, so when requests start coming from the cache server there is no problem.

ISPs which transparently cache leased-line customers are the most likely to have problems with IP-authenticating servers. If you are phasing transparency in for such an ISP, you must make sure that your customers know all the implications. They must know how to refresh pages (and who to tell if they find such out-of-date pages, so that the Squid refresh rules can be changed), and how the source IP address is going to change. You must not simply install the transparent cache and hope for the best!

The Transparent Caching Process

Let's look at what happens when you use transparency. First, though, you need to know something of what happens to IP packets at the ethernet level.

Some Routing Basics

An ethernet IP packet contains four addresses:

When a host wants to communicate with a machine that isn't on the local network, it uses a smart router to find the path to that network. When the client wants to send a packet through a router, the client sets the destination mac address of the packet to the router's interface, and sets the IP destination address to the required end host. It's important to know that the destination IP address of the packet isn't set to the router's IP address, only the mac address is changed. When a router accepts a packet, it decides which host to forward it to, based on it's routing tables. The router then sets the destination mac address of the packet to the next-hop router's ethernet address, and sends the packet to that machine. The remote host then repeats this process: if it's the destination machine, it uses the packet, but if it's another router, it will try and move the packet closer to it's final destination.

Packet Flow with Transparent Caches

Transparent caches essentially look out for TCP connections destined for port 80. The cache server will intercept these packets, convert them to a standard TCP stream and pass them to Squid. When Squid sends reply data to the client, the Operating System fakes the source address of the packets, so that the client believes it is connected to the server that it originally sent the request to.

You can't simply plug a transparent cache into the network and get it to transparently cache pages. The cache server needs to be in a position where it can fake the reply packets (without the real server interrupting the conversation and confusing things.) The server needs to be the gateway to the outside world.

Let's look at the simplest transparent cache setup. The client machine ( treats the cache server's internal ( interface as it's default gateway. This way, all packets arrive on the cache server before they reach the rest of the Internet. The filter looks for port 80 packets, and passes them to Squid, but allows all other packets to be passed to the routing layer, which passes the packets to the router's IP (

Once the connection is established, Squid needs to communicate with the client. Squid doesn't do any strange packet assembly: that's left to the transparency layer. When Squid sends reply data to the client, the kernel automatically changes the packet's from address, so it appears to the client that the server is just routing the requests from the outside world. When Squid connects to the remote server, however, the connect comes from the external interface of the cache server (, in the example.) This is where IP-authentication breaks: since the request is coming from the cache (rather than the client's real address (

Effectively, we need to get four things right to get transparency right:

Chapter 10. Not Yet Done: Squid Config files and options

3.2: Squid Command Line Options
3.2.1: Help
To get a complete list of Squid's command-line options, with a short
description of each option, use the '-h' option.
3.2.2: HTTP Port
Option: -a
Format: -a port number
Example: squid -a 3128
Squid will normally accept incoming HTTP requests on the port specified in
the squid.conf file with the http_port tag. If you wish to override the tag
for some reason, you can use the '-a' option.
3.2.3: Debug Information
Option: -d
Format: -d debug level value
Example: squid -d 3
By default Squid only logs fatal errors to the screen, logging all other
errors to the cache.log file.
If you wish to log more information (for example debugging information,
rather than only errors)
The "-d" option allows you to increase the amount of debug information
logged to the screen. If squid is started from your startup scripts, then
this output will appear on the console of the machine. If started from a
remote login, this output will be written to the screen of your remote
3.2.4: Config file
Option: -f
Format: -f path
Example: squid -f /usr/local/etc/squid.conf
This option allows you to specify a different path to the squid config file.
When installing a binary version of squid, the default path to the
squid.conf file may be inappropriate to your system.
If you wish to test a different version of the config file, but wish to be
able to revert to the previous config file in a hurry, you can use this
option to refer to a different config file. To change back to the other
config file you just have to restart Squid without this option.
3.2.5: Signaling a running Squid
Option: -k
Format: -k action
Example: squid -k rotate
You can communicate with a running copy of Squid by sending it signals.
These signals will cause Squid to perform maintenance functions, doing
things like reloading the config file, rotating the logs (for analysis) and
so forth.
On some operating systems certain signals are reserved. The threads library
on Linux, for example, uses the SIGUSR1 and SIGUSR2 signals for thread
Sending the wrong signal to a running Squid is easy, and can have
unfortunate consequences.
This option allows you to use descriptive options to send a running Squid
signals, creating a standardized cross-platform user interface.
Tag: reconfigure
Action: Reloads the squid.conf file.
It's important to note that when Squid re-reads this file it
closes all current connections, which means that clients that were
downloading files will be cut off mid-download. You should only
schedule reloads for after-hours, when their impact is minimal.
Tag: rotate
Action: Rotates the cache.log and access.log files
Cache log files get very large. To stop the log files using
up all your disk space you should rotate the logs daily.
The squid.conf logfile_rotate option sets the maximum number of
rotated logs that you wish to keep.
The most common use of this action is to rotate the logs just
before logfile analysis (see Chapter 10). A crontab signals
the rotation, sleeps for a short time, and then calls the logfile
analysis program.
Tags: shutdown, interrupt
Action: Closes current connections, writes index and exits
Squid keeps an index of cache objects in memory. When you wish to
shutdown Squid you should use this option, rather than simply
killing Squid. Shutting down Squid can take a short while, while it
writes the object index to disk. Squid writes to the cache.log
file while it shuts down, indicating how many objects it has
written to the index.
Both the shutdown and interrupt tag have the same effect. (?why
I thing it's since there isn't a "kill" command for NT?)
Tag: kill
Action: Kills the Squid process
The kill tag should only be used if shutdown or interrupt have
no effect. Using this tag will kill Squid without giving it a
chance to write the cache index file, causing a slow rebuild on the
next start.
Tag: debug
Action: Turns on maximum debugging
At times it is useful to see exactly what the running copy of Squid
is doing. Using the debug option will turn maximum logging on for
the main Squid process. The output is very verbose, and with a
heavily loaded cache can consume megabytes of disk space, so use
this only on a lightly loaded cache, and for small periods of time.
Tag: check
Action: Prints an error message if Squid isn't running
Using this tag causes a 'kill -0' signal to the running copy of
Squid. This doesn't do anything to the running process, other than
check that it exists (and that the user running the command has
permission to send signals to the process).
3.2.5: Logging to syslog
Option: -s
Format: -s
Example: squid -s
Squid normally logs events and debug information to a special file,
normally stored in "/usr/local/squid/logs/cache.log".
In some environments you may wish for events to be logged to central "log
server", using syslog.
Turning on this flag will 
are not logged to syslog. Logs of client accesses are stored in
the file "/usr/local/squid/logs/access.log"
Squid is designed with the ability to store millions of objects. Given that
many operating systems have a limit on file size it's not feasible for a
cross platform program like Squid to store all objects in one file, though
there are patches to allow users to create squid stores on large files or
on raw devices.
If you run a news server you will probably have an idea of how slow it is
to do a directory listing of a directory with hundreds of thousands of
files in it. On almost all filesystems there is a linear slowdown as more
files are added to a directory. This rules out the other option, creating
unique filenames and storing them all in one directory.
Squid uses a hierarchy of directories for file storage. The default setup
creates 16 first-tier directories. Each one of these directories then
contains 256 second-tier directories. Files are only stored in the
second-tier directories. This

Chapter 11. Overall Layout (for writers)

When originally working on this document, I (Oskar Pearson) put together an overall structure text file, to help me decide where all the info would go, and how we would introduce the reader to squid in a structured, useful fashion. You are welcome to change the ordering, and add and remove from it as you will. The numbers below do not directly match with the final structure, nor do the numbers used match the numbers used in the free-release version of the document. The "X pages" values may be freely ignored; they were for the original proposal for the (paper) book.

You are welcome to change the overall document structure as you wish, and to add new sections as you will. If you do, though, please bear in mind that someone coming into the editing process may benefit from having an "overall structure" in which to place their contribution. If you alter the document structure, please change this section of the document to match. If you find differences between the flow of concepts in this area and the text (entirely possible), you should probably ignore this section, and (preferably) update this to match the real document.

I) Installation
    When installing Squid the first step is to get it up and running
    on a test machine. This allows the user to get familiar with
    Squid's basic setup and feel that they are progressing towards
    something tangible (rather than slogging through the whole
    book before actually getting Squid running). Only the very
    basics of the config file are going to be covered.

Chapter 1) Introduction to Squid terminology and technology
    1.1) What Squid is
    1.1.1) Why cache
    1.2) What Squid is not
    1.3) Supported Internet Server protocols
    1.4) Inter-Cache communication
    1.4.1) Hierarchy terminology
    1.4.2) Inter-Cache protocols
    1.5) Operating Systems

Chapter 2)
    2.1) Advanced Planning:
    2.1.1) Hardware requirements:
    2.2) Operating System:
    2.2.1) Use the OS that you have experience in
    2.2.2) All examples will be generic.
    2.2.3) need a compiler.
    2.3) System setup
    2.3.1) The default Squid directory structure
    2.3.2) Creation of the squid user and group
    (includes permissions etc)
    2.4) Working with precompiled binaries
    2.4.1) precompiled binaries
    2.4.2) Trusted sources of binaries
    2.5) Source compilation
    2.5.1) Recommended compilation tools
    2.5.2) Compilation configuration options
    2.5.3) compilation: make all; make install

Chapter 3) Introduction to the configuration file:
    Only the very basics of the config file are covered. This
    allows people to get Squid running as soon as they can.

    3.1) note on RCS
    3.2) The configuration file:
    3.2.1) HTTP port
    3.2.2) Communicating with other proxy servers Basic cache hierarchy terminology Proxy-level firewall Packet-filter firewall Source/Destination IP and Port pairs
    3.2.3) Cache Store location Disk space allocation (? move to
    3.2.4) FTP login information
    3.2.5) acl, http_access create a basic acl that denies
   everything but one address range Intranet access with parents
    3.2.6) cache_mgr
    3.2.7) cache_effective_group

Chapter 4) Starting and Running Squid (15 pages)
    4.1) Running Squid for the first time
    4.1.1) Permissions
    - on each ~squid/* directory
    4.1.2) Creating cache directories Problems creating Swap Directories
    - problems:
    not root
    squid user id doesn't exist
    squid user doesn't have write to
    cache dir
    squid user doesn't have read/exec
    to a directory up the tree
    4.2) Running Squid
    3.2.1) What is expected in cache.log
    4.3) Testing the cache with the included client
    4.3.1) checking if Internet works
    4.3.2) checking if intranet works (if configured with a parent)
    4.3.3) Checking Access.log for hits vs misses
   Include basic fields
    4.4) Addition to startup files
    (? check NT ?)

Chapter 5) Client configuration: (24 pages)
    Include some screen shots of the configuration menus
    5.1) Basic client configuration.
    5.1.1) Netscape
    5.1.2) Internet Explorer
    5.1.3) Unix environment variables
    (Important for both lynx and for wget - for
    prefetching pages)
    5.2) Client cache-specific modifications
    5.3) Testing client access
    5.4) Setting clients to use LOCAL caches
 5.4.1) CARP
 5.4.2) Autoconfigs
 5.4.3) Future directions DNS destination selection based on Roaming ability will help Transparency (see 11.1)

II) Integration
    By this point Squid should be installed with a minimum working

    This section covers changing cache setup to suit the local network

    This section covers Access Control, Refresh patterns and Cache-peer
    relationships. These are the painful sections of the setup.

    This section also goes through the options in the config file
    that haven't been covered. This is essentially a 'reference
    guide' to the config options.

Chapter 6) ACLs: (38 pages)
    Each of these includes a short example that shows how
    they work. At the end of the Chapter there is a nice long
    complex ACL list that should suit almost everyone.

    6.1) Introduction to ACLs
    6.1.1) ACL lines vs Operator lines
    6.1.2) How decisions work
    6.2) Data specification:
    6.2.1) regular expressions
    6.2.2) IP address range specifications
    6.2.3) AS numbers
    6.2.4) putting the data in files
    6.3) types of acl lines:
 Works through all the acl types. (src, srcdomain, dst,
    dstdomain etc)
    - must include info on "no_cache", specifically
    6.4) Delay classes 
    6.5) SNMP configuration
    6.5) The default acl set
    include info on why the SSL setup is the way that it is,
    and information on the echo/chargen ports

Chapter 7) Hierarchies: (42 pages)
    7.1) Inter-cache communication protocols
 How each one is suited to specific
 circumstances. Compatability notes
 (with other programs) are included.
    7.1.1) ICP
    7.1.2) Digests
    7.1.3) HTCP
    7.1.4) CARP
    7.2) Various types of hierarchy structures
 are covered:
     7.2.1) The Tree stucture
     7.2.2) Load balancing peer system
     7.2.3) True distributed system
    7.3) Configuring Squid to talk to other caches
    7.3.1) The cache_peer config option
   All options are covered with examples
    7.3.2) cache_peer_domain config option
    7.3.3) miss_access acl line

Chapter 8) Accelerator mode (11 pages)
    (? I haven't use accelerator mode - I am using Miguel a.l. Paraz's
       page in the Squid Documentation as a guide ?)
    8.1) Intro - why use this mode
    8.1.1) performance
    8.1.2) security
    8.2) Types of accelerator mode
    8.2.1) Virtual mode
    (note on security problems)
    8.2.2) Host header
    8.3) Options
    8.3.1) http_port
    8.3.2) httpd_accel_host
    8.3.3) httpd_accel_port
    8.3.4) httpd_accel_with_proxy
    8.3.5) httpd_accel_uses_host_header
    8.4) Reverse caching using accelerator mode on the return
 path of an International link
    See Transparency

Chapter 9) Transparency: (24 pages)
    9.1) TCP basics
    9.2) Operating System function
    9.3) Squid 'accept' destination sensing
    9.4) Special ACLs to stop loops
    9.5) FTP transparency problems
    9.6) Routing the actual TCP packets to Squid
    9.7) Changing hierarchies to work with transparency

Chapter 10) The config file and Squid options (48 pages)

    The options list doesn't really belong in section (I). I am,
    instead going to cover it here. Also cover the options to 'client'.

    This covers ALL the tags in the config file. Where
    the tag has been covered already it refers people to
    that section of the book.

    Arranged in alphabetical order.

III) Maintainence and Site-Specific Optimization
    Covers the further development of your cache setup. This covers
    both maintainence and specialized setups (like transparent caches)

Chapter 11) Refresh Patterns: (24 pages)
    11.1) distribution of file types
    (gifs vs jpg vs html)
    11.2) distribution of protocols
    11.3) Server-Sent Header fields
    11.3.1) Work through the types of headers
    11.3.2) meta-tags
    11.4) Client-Sent Header fields
    11.4.1) If-Modified-Since Requests
    11.4.2) Refresh button
    11.5) refresh_pattern tag
 First match selection. Describes order of
 checking each of the fields.

Chapter 12) Cache analysis (24 pages)
      This section covers disadvantages and advantages
      of the various types of cache performance/savings
      analysis systems
    12.1) access.log fields
    12.2) Simple Network Management Protocol (SNMP)
    configuring, access control, multiple servers,
    multiple agent configurations, understanding
    results. Shew!
    12.3) Cache-specific analysis using a squid analysis
    12.4) The cachemgr.cgi script
    Using the output (eg LRU values)
    for deciding when to buy more disk space etc
    12.5) Using a cache-query-tool
    12.6) Using your results
    Graphing response times over the months, for example.

Chapter 13) Standby procedures: (15 pages)
    13.1) Hardware failure
    13.1.1) Standby machines
    13.1.2) DNS modification
    13.1.3) Automatic configuration
    13.2) Software failure
    We need lots of info on 'vmstat', 'iostat', strace -T,
    (and other stuff like that) here.
    queued DNS queries
    DNS response times
    queued username/password authentication
    page faults:
    13.2.2) Consistent crashing
    - filehandles
    - memory
    - all dnsservers busy
    - slow!
    - latency of local request,
      comparing with "client" through
      cache and without it.

Chapter 14) Future directions: (18 pages)
    14.1) Wide ranging use of Skycache
    14.2) Wide ranging use of transparency
    14.3) Very heavily used parents
  For example at Exchange Points
    14.4) compression between server and client - like the berkely

Chapter 12. GNU Free Documentation License

GNU Free Documentation License

GNU Free Documentation License
   Version 1.1, March 2000

 Copyright (C) 2000  Free Software Foundation, Inc.
     59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 Everyone is permitted to copy and distribute verbatim copies
 of this license document, but changing it is not allowed.


The purpose of this License is to make a manual, textbook, or other
written document "free" in the sense of freedom: to assure everyone
the effective freedom to copy and redistribute it, with or without
modifying it, either commercially or noncommercially.  Secondarily,
this License preserves for the author and publisher a way to get
credit for their work, while not being considered responsible for
modifications made by others.

This License is a kind of "copyleft", which means that derivative
works of the document must themselves be free in the same sense.  It
complements the GNU General Public License, which is a copyleft
license designed for free software.

We have designed this License in order to use it for manuals for free
software, because free software needs free documentation: a free
program should come with manuals providing the same freedoms that the
software does.  But this License is not limited to software manuals;
it can be used for any textual work, regardless of subject matter or
whether it is published as a printed book.  We recommend this License
principally for works whose purpose is instruction or reference.


This License applies to any manual or other work that contains a
notice placed by the copyright holder saying it can be distributed
under the terms of this License.  The "Document", below, refers to any
such manual or work.  Any member of the public is a licensee, and is
addressed as "you".

A "Modified Version" of the Document means any work containing the
Document or a portion of it, either copied verbatim, or with
modifications and/or translated into another language.

A "Secondary Section" is a named appendix or a front-matter section of
the Document that deals exclusively with the relationship of the
publishers or authors of the Document to the Document's overall subject
(or to related matters) and contains nothing that could fall directly
within that overall subject.  (For example, if the Document is in part a
textbook of mathematics, a Secondary Section may not explain any
mathematics.)  The relationship could be a matter of historical
connection with the subject or with related matters, or of legal,
commercial, philosophical, ethical or political position regarding

The "Invariant Sections" are certain Secondary Sections whose titles
are designated, as being those of Invariant Sections, in the notice
that says that the Document is released under this License.

The "Cover Texts" are certain short passages of text that are listed,
as Front-Cover Texts or Back-Cover Texts, in the notice that says that
the Document is released under this License.

A "Transparent" copy of the Document means a machine-readable copy,
represented in a format whose specification is available to the
general public, whose contents can be viewed and edited directly and
straightforwardly with generic text editors or (for images composed of
pixels) generic paint programs or (for drawings) some widely available
drawing editor, and that is suitable for input to text formatters or
for automatic translation to a variety of formats suitable for input
to text formatters.  A copy made in an otherwise Transparent file
format whose markup has been designed to thwart or discourage
subsequent modification by readers is not Transparent.  A copy that is
not "Transparent" is called "Opaque".

Examples of suitable formats for Transparent copies include plain
ASCII without markup, Texinfo input format, LaTeX input format, SGML
or XML using a publicly available DTD, and standard-conforming simple
HTML designed for human modification.  Opaque formats include
PostScript, PDF, proprietary formats that can be read and edited only
by proprietary word processors, SGML or XML for which the DTD and/or
processing tools are not generally available, and the
machine-generated HTML produced by some word processors for output
purposes only.

The "Title Page" means, for a printed book, the title page itself,
plus such following pages as are needed to hold, legibly, the material
this License requires to appear in the title page.  For works in
formats which do not have any title page as such, "Title Page" means
the text near the most prominent appearance of the work's title,
preceding the beginning of the body of the text.


You may copy and distribute the Document in any medium, either
commercially or noncommercially, provided that this License, the
copyright notices, and the license notice saying this License applies
to the Document are reproduced in all copies, and that you add no other
conditions whatsoever to those of this License.  You may not use
technical measures to obstruct or control the reading or further
copying of the copies you make or distribute.  However, you may accept
compensation in exchange for copies.  If you distribute a large enough
number of copies you must also follow the conditions in section 3.

You may also lend copies, under the same conditions stated above, and
you may publicly display copies.


If you publish printed copies of the Document numbering more than 100,
and the Document's license notice requires Cover Texts, you must enclose
the copies in covers that carry, clearly and legibly, all these Cover
Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on
the back cover.  Both covers must also clearly and legibly identify
you as the publisher of these copies.  The front cover must present
the full title with all words of the title equally prominent and
visible.  You may add other material on the covers in addition.
Copying with changes limited to the covers, as long as they preserve
the title of the Document and satisfy these conditions, can be treated
as verbatim copying in other respects.

If the required texts for either cover are too voluminous to fit
legibly, you should put the first ones listed (as many as fit
reasonably) on the actual cover, and continue the rest onto adjacent

If you publish or distribute Opaque copies of the Document numbering
more than 100, you must either include a machine-readable Transparent
copy along with each Opaque copy, or state in or with each Opaque copy
a publicly-accessible computer-network location containing a complete
Transparent copy of the Document, free of added material, which the
general network-using public has access to download anonymously at no
charge using public-standard network protocols.  If you use the latter
option, you must take reasonably prudent steps, when you begin
distribution of Opaque copies in quantity, to ensure that this
Transparent copy will remain thus accessible at the stated location
until at least one year after the last time you distribute an Opaque
copy (directly or through your agents or retailers) of that edition to
the public.

It is requested, but not required, that you contact the authors of the
Document well before redistributing any large number of copies, to give
them a chance to provide you with an updated version of the Document.


You may copy and distribute a Modified Version of the Document under
the conditions of sections 2 and 3 above, provided that you release
the Modified Version under precisely this License, with the Modified
Version filling the role of the Document, thus licensing distribution
and modification of the Modified Version to whoever possesses a copy
of it.  In addition, you must do these things in the Modified Version:

A. Use in the Title Page (and on the covers, if any) a title distinct
   from that of the Document, and from those of previous versions
   (which should, if there were any, be listed in the History section
   of the Document).  You may use the same title as a previous version
   if the original publisher of that version gives permission.
B. List on the Title Page, as authors, one or more persons or entities
   responsible for authorship of the modifications in the Modified
   Version, together with at least five of the principal authors of the
   Document (all of its principal authors, if it has less than five).
C. State on the Title page the name of the publisher of the
   Modified Version, as the publisher.
D. Preserve all the copyright notices of the Document.
E. Add an appropriate copyright notice for your modifications
   adjacent to the other copyright notices.
F. Include, immediately after the copyright notices, a license notice
   giving the public permission to use the Modified Version under the
   terms of this License, in the form shown in the Addendum below.
G. Preserve in that license notice the full lists of Invariant Sections
   and required Cover Texts given in the Document's license notice.
H. Include an unaltered copy of this License.
I. Preserve the section entitled "History", and its title, and add to
   it an item stating at least the title, year, new authors, and
   publisher of the Modified Version as given on the Title Page.  If
   there is no section entitled "History" in the Document, create one
   stating the title, year, authors, and publisher of the Document as
   given on its Title Page, then add an item describing the Modified
   Version as stated in the previous sentence.
J. Preserve the network location, if any, given in the Document for
   public access to a Transparent copy of the Document, and likewise
   the network locations given in the Document for previous versions
   it was based on.  These may be placed in the "History" section.
   You may omit a network location for a work that was published at
   least four years before the Document itself, or if the original
   publisher of the version it refers to gives permission.
K. In any section entitled "Acknowledgements" or "Dedications",
   preserve the section's title, and preserve in the section all the
   substance and tone of each of the contributor acknowledgements
   and/or dedications given therein.
L. Preserve all the Invariant Sections of the Document,
   unaltered in their text and in their titles.  Section numbers
   or the equivalent are not considered part of the section titles.
M. Delete any section entitled "Endorsements".  Such a section
   may not be included in the Modified Version.
N. Do not retitle any existing section as "Endorsements"
   or to conflict in title with any Invariant Section.

If the Modified Version includes new front-matter sections or
appendices that qualify as Secondary Sections and contain no material
copied from the Document, you may at your option designate some or all
of these sections as invariant.  To do this, add their titles to the
list of Invariant Sections in the Modified Version's license notice.
These titles must be distinct from any other section titles.

You may add a section entitled "Endorsements", provided it contains
nothing but endorsements of your Modified Version by various
parties--for example, statements of peer review or that the text has
been approved by an organization as the authoritative definition of a

You may add a passage of up to five words as a Front-Cover Text, and a
passage of up to 25 words as a Back-Cover Text, to the end of the list
of Cover Texts in the Modified Version.  Only one passage of
Front-Cover Text and one of Back-Cover Text may be added by (or
through arrangements made by) any one entity.  If the Document already
includes a cover text for the same cover, previously added by you or
by arrangement made by the same entity you are acting on behalf of,
you may not add another; but you may replace the old one, on explicit
permission from the previous publisher that added the old one.

The author(s) and publisher(s) of the Document do not by this License
give permission to use their names for publicity for or to assert or
imply endorsement of any Modified Version.


You may combine the Document with other documents released under this
License, under the terms defined in section 4 above for modified
versions, provided that you include in the combination all of the
Invariant Sections of all of the original documents, unmodified, and
list them all as Invariant Sections of your combined work in its
license notice.

The combined work need only contain one copy of this License, and
multiple identical Invariant Sections may be replaced with a single
copy.  If there are multiple Invariant Sections with the same name but
different contents, make the title of each such section unique by
adding at the end of it, in parentheses, the name of the original
author or publisher of that section if known, or else a unique number.
Make the same adjustment to the section titles in the list of
Invariant Sections in the license notice of the combined work.

In the combination, you must combine any sections entitled "History"
in the various original documents, forming one section entitled
"History"; likewise combine any sections entitled "Acknowledgements",
and any sections entitled "Dedications".  You must delete all sections
entitled "Endorsements."


You may make a collection consisting of the Document and other documents
released under this License, and replace the individual copies of this
License in the various documents with a single copy that is included in
the collection, provided that you follow the rules of this License for
verbatim copying of each of the documents in all other respects.

You may extract a single document from such a collection, and distribute
it individually under this License, provided you insert a copy of this
License into the extracted document, and follow this License in all
other respects regarding verbatim copying of that document.


A compilation of the Document or its derivatives with other separate
and independent documents or works, in or on a volume of a storage or
distribution medium, does not as a whole count as a Modified Version
of the Document, provided no compilation copyright is claimed for the
compilation.  Such a compilation is called an "aggregate", and this
License does not apply to the other self-contained works thus compiled
with the Document, on account of their being thus compiled, if they
are not themselves derivative works of the Document.

If the Cover Text requirement of section 3 is applicable to these
copies of the Document, then if the Document is less than one quarter
of the entire aggregate, the Document's Cover Texts may be placed on
covers that surround only the Document within the aggregate.
Otherwise they must appear on covers around the whole aggregate.


Translation is considered a kind of modification, so you may
distribute translations of the Document under the terms of section 4.
Replacing Invariant Sections with translations requires special
permission from their copyright holders, but you may include
translations of some or all Invariant Sections in addition to the
original versions of these Invariant Sections.  You may include a
translation of this License provided that you also include the
original English version of this License.  In case of a disagreement
between the translation and the original English version of this
License, the original English version will prevail.


You may not copy, modify, sublicense, or distribute the Document except
as expressly provided for under this License.  Any other attempt to
copy, modify, sublicense or distribute the Document is void, and will
automatically terminate your rights under this License.  However,
parties who have received copies, or rights, from you under this
License will not have their licenses terminated so long as such
parties remain in full compliance.


The Free Software Foundation may publish new, revised versions
of the GNU Free Documentation License from time to time.  Such new
versions will be similar in spirit to the present version, but may
differ in detail to address new problems or concerns.  See

Each version of the License is given a distinguishing version number.
If the Document specifies that a particular numbered version of this
License "or any later version" applies to it, you have the option of
following the terms and conditions either of that specified version or
of any later version that has been published (not as a draft) by the
Free Software Foundation.  If the Document does not specify a version
number of this License, you may choose any version ever published (not
as a draft) by the Free Software Foundation.

ADDENDUM: How to use this License for your documents

To use this License in a document you have written, include a copy of
the License in the document and put the following copyright and
license notices just after the title page:

      Copyright (c)  YEAR  YOUR NAME.
      Permission is granted to copy, distribute and/or modify this document
      under the terms of the GNU Free Documentation License, Version 1.1
      or any later version published by the Free Software Foundation;
      with the Invariant Sections being LIST THEIR TITLES, with the
      Front-Cover Texts being LIST, and with the Back-Cover Texts being LIST.
      A copy of the license is included in the section entitled "GNU
      Free Documentation License".

If you have no Invariant Sections, write "with no Invariant Sections"
instead of saying which ones are invariant.  If you have no
Front-Cover Texts, write "no Front-Cover Texts" instead of
"Front-Cover Texts being LIST"; likewise for Back-Cover Texts.

If your document contains nontrivial examples of program code, we
recommend releasing these examples in parallel under your choice of
free software license, such as the GNU General Public License,
to permit their use in free software.