7. The Main Configuration File

The directory where the other configuration files are located

Path name

/etc/privoxy (Unix) or Privoxy installation dir (Windows)

Mandatory

No trailing "/", please

When development goes modular and multi-user, the blocker, filter, and per-user config will be stored in subdirectories of "confdir". For now, the configuration directory structure is flat, except for confdir/templates, where the HTML templates for CGI output reside (e.g. Privoxy's 404 error page).

The directory where all logging takes place (i.e. where logfile and jarfile are located)

Path name

/var/log/privoxy (Unix) or Privoxy installation dir (Windows)

Mandatory

No trailing "/", please

The actions file(s) to use

File name, relative to confdir, without the .action suffix

No actions are taken at all. Simple neutral proxying.

Multiple actionsfile lines are permitted, and are in fact recommended!

The default values include standard.action, which is used for internal purposes and should be loaded, default.action, which is the "main" actions file maintained by the developers, and user.action, where you can make your personal additions.

Actions files are where all the per site and per URL configuration is done for ad blocking, cookie management, privacy considerations, etc. There is no point in using Privoxy without at least one actions file.

The filter file to use

File name, relative to confdir

default.filter (Unix) or default.filter.txt (Windows)

No textual content filtering takes place, i.e. all +filter{name} actions in the actions files are turned neutral.

The filter file contains content modification rules that use regular expressions. These rules permit powerful changes on the content of Web pages, e.g., you could disable your favorite JavaScript annoyances, re-write the actual displayed text, or just have some fun replacing "Microsoft" with "MicroSuck" wherever it appears on a Web page.

The +filter{name} actions rely on the relevant filter (name) to be defined in the filter file!

A pre-defined filter file called default.filter that contains a bunch of handy filters for common problems is included in the distribution. See the section on the filter action for a list.

The log file to use

File name, relative to logdir

logfile (Unix) or privoxy.log (Windows)

No log file is used, all log messages go to the console (STDERR).

The windows version will additionally log to the console.

The logfile is where all logging and error messages are written. The level of detail and number of messages are set with the debug option (see below). The logfile can be useful for tracking down a problem with Privoxy (e.g., it's not blocking an ad you think it should block) but in most cases you probably will never look at it.

Your logfile will grow indefinitely, and you will probably want to periodically remove it. On Unix systems, you can do this with a cron job (see "man cron"). For Red Hat, a logrotate script has been included.

On SuSE Linux systems, you can place a line like "/var/log/privoxy.* +1024k 644 nobody.nogroup" in /etc/logfiles, with the effect that cron.daily will automatically archive, gzip, and empty the log, when it exceeds 1M size.

Any log files must be writable by whatever user Privoxy is being run as (default on UNIX, user id is "privoxy").

The file to store intercepted cookies in

File name, relative to logdir

jarfile (Unix) or privoxy.jar (Windows)

Intercepted cookies are not stored at all.

The jarfile may grow to ridiculous sizes over time.

The trust file to use

File name, relative to confdir

Unset (commented out). When activated: trust (Unix) or trust.txt (Windows)

The whole trust mechanism is turned off.

The trust mechanism is an experimental feature for building white-lists and should be used with care. It is NOT recommended for the casual user.

If you specify a trust file, Privoxy will only allow access to sites that are named in the trustfile. You can also mark sites as trusted referrers (with +), with the effect that access to untrusted sites will be granted, if a link from a trusted referrer was used. The link target will then be added to the "trustfile". Possible applications include limiting Internet access for children.

If you use + operator in the trust file, it may grow considerably over time.

Location of the Privoxy User Manual.

A fully qualified URI

Unset

http://www.privoxy.org/version/user-manual/ will be used, where version is the Privoxy version.

The User Manual URI is used for help links from some of the internal CGI pages. The manual itself is normally packaged with the binary distributions, so you probably want to set this to a locally installed copy. For multi-user setups, you could provide a copy on a local webserver for all your users and use the corresponding URL here.

Examples:

Unix, in local filesystem:

user-manual  file:///usr/share/doc/privoxy-3.0.0/user-manual/

Any platform, on local webserver (called "local-webserver"):

user-manual  http://local-webserver/privoxy-user-manual/

A URL to be displayed in the error page that users will see if access to an untrusted page is denied.

URL

Two example URL are provided

No links are displayed on the "untrusted" error page.

The value of this option only matters if the experimental trust mechanism has been activated. (See trustfile above.)

If you use the trust mechanism, it is a good idea to write up some on-line documentation about your trust policy and to specify the URL(s) here. Use multiple times for multiple URLs.

The URL(s) should be added to the trustfile as well, so users don't end up locked out from the information on why they were locked out in the first place!

An email address to reach the proxy administrator.

Email address

Unset

No email address is displayed on error pages and the CGI user interface.

If both admin-address and proxy-info-url are unset, the whole "Local Privoxy Support" box on all generated pages will not be shown.

A URL to documentation about the local Privoxy setup, configuration or policies.

URL

Unset

No link to local documentation is displayed on error pages and the CGI user interface.

If both admin-address and proxy-info-url are unset, the whole "Local Privoxy Support" box on all generated pages will not be shown.

This URL shouldn't be blocked ;-)

Key values that determine what information gets logged to the logfile.

Integer values

12289 (i.e.: URLs plus informational and warning messages)

Nothing gets logged.

The available debug levels are:

debug 1 # show each GET/POST/CONNECT request debug 2 # show each connection status debug 4 # show I/O status debug 8 # show header parsing debug 16 # log all data into the logfile debug 32 # debug force feature debug 64 # debug regular expression filter debug 128 # debug fast redirects debug 256 # debug GIF de-animation debug 512 # Common Log Format debug 1024 # debug kill pop-ups debug 2048 # CGI user interface debug 4096 # Startup banner and warnings. debug 8192 # Non-fatal errors

To select multiple debug levels, you can either add them or use multiple debug lines.

A debug level of 1 is informative because it will show you each request as it happens. 1, 4096 and 8192 are highly recommended so that you will notice when things go wrong. The other levels are probably only of interest if you are hunting down a specific problem. They can produce a hell of an output (especially 16).

The reporting of fatal errors (i.e. ones which crash Privoxy) is always on and cannot be disabled.

If you want to use CLF (Common Log Format), you should set "debug 512" ONLY and not enable anything else.

Whether to run only one server thread

None

Unset

Multi-threaded (or, where unavailable: forked) operation, i.e. the ability to serve multiple requests simultaneously.

This option is only there for debug purposes and you should never need to use it. It will drastically reduce performance.

The IP address and TCP port on which Privoxy will listen for client requests.

[IP-Address]:Port

127.0.0.1:8118

Bind to 127.0.0.1 (localhost), port 8118. This is suitable and recommended for home users who run Privoxy on the same machine as their browser.

You will need to configure your browser(s) to this proxy address and port.

If you already have another service running on port 8118, or if you want to serve requests from other machines (e.g. on your local network) as well, you will need to override the default.

If you leave out the IP address, Privoxy will bind to all interfaces (addresses) on your machine and may become reachable from the Internet. In that case, consider using access control lists (ACL's, see below), and/or a firewall.

If you open Privoxy to untrusted users, you will also want to turn off the enable-edit-actions and enable-remote-toggle options!

Suppose you are running Privoxy on a machine which has the address 192.168.0.1 on your local private network (192.168.0.0) and has another outside connection with a different address. You want it to serve requests from inside only:

listen-address 192.168.0.1:8118

Initial state of "toggle" status

1 or 0

1

Act as if toggled on

If set to 0, Privoxy will start in "toggled off" mode, i.e. behave like a normal, content-neutral proxy where all ad blocking, filtering, etc are disabled. See enable-remote-toggle below. This is not really useful anymore, since toggling is much easier via the web interface than via editing the conf file.

The windows version will only display the toggle icon in the system tray if this option is present.

Whether or not the web-based toggle feature may be used

0 or 1

1

The web-based toggle feature is disabled.

When toggled off, Privoxy acts like a normal, content-neutral proxy, i.e. it acts as if none of the actions applied to any URL.

For the time being, access to the toggle feature can not be controlled separately by "ACLs" or HTTP authentication, so that everybody who can access Privoxy (see "ACLs" and listen-address above) can toggle it for all users. So this option is not recommended for multi-user environments with untrusted users.

Note that you must have compiled Privoxy with support for this feature, otherwise this option has no effect.

Whether or not the web-based actions file editor may be used

0 or 1

1

The web-based actions file editor is disabled.

For the time being, access to the editor can not be controlled separately by "ACLs" or HTTP authentication, so that everybody who can access Privoxy (see "ACLs" and listen-address above) can modify its configuration for all users. So this option is not recommended for multi-user environments with untrusted users.

Note that you must have compiled Privoxy with support for this feature, otherwise this option has no effect.

Who can access what.

src_addr[/src_masklen] [dst_addr[/dst_masklen]]

Where src_addr and dst_addr are IP addresses in dotted decimal notation or valid DNS names, and src_masklen and dst_masklen are subnet masks in CIDR notation, i.e. integer values from 2 to 30 representing the length (in bits) of the network address. The masks and the whole destination part are optional.

Unset

Don't restrict access further than implied by listen-address

Access controls are included at the request of ISPs and systems administrators, and are not usually needed by individual users. For a typical home user, it will normally suffice to ensure that Privoxy only listens on the localhost (127.0.0.1) or internal (home) network address by means of the listen-address option.

Please see the warnings in the FAQ that this proxy is not intended to be a substitute for a firewall or to encourage anyone to defer addressing basic security weaknesses.

Multiple ACL lines are OK. If any ACLs are specified, then the Privoxy talks only to IP addresses that match at least one permit-access line and don't match any subsequent deny-access line. In other words, the last match wins, with the default being deny-access.

If Privoxy is using a forwarder (see forward below) for a particular destination URL, the dst_addr that is examined is the address of the forwarder and NOT the address of the ultimate target. This is necessary because it may be impossible for the local Privoxy to determine the IP address of the ultimate target (that's often what gateways are used for).

You should prefer using IP addresses over DNS names, because the address lookups take time. All DNS names must resolve! You can not use domain patterns like "*.org" or partial domain names. If a DNS name resolves to multiple IP addresses, only the first one is used.

Denying access to particular sites by ACL may have undesired side effects if the site in question is hosted on a machine which also hosts other sites.

Explicitly define the default behavior if no ACL and listen-address are set: "localhost" is OK. The absence of a dst_addr implies that all destination addresses are OK:

permit-access localhost

Allow any host on the same class C subnet as www.privoxy.org access to nothing but www.example.com:

permit-access www.privoxy.org/24 www.example.com/32

Allow access from any host on the 26-bit subnet 192.168.45.64 to anywhere, with the exception that 192.168.45.73 may not access www.dirty-stuff.example.com:

permit-access 192.168.45.64/26 deny-access 192.168.45.73 www.dirty-stuff.example.com

Maximum size of the buffer for content filtering.

Size in Kbytes

4096

Use a 4MB (4096 KB) limit.

For content filtering, i.e. the +filter and +deanimate-gif actions, it is necessary that Privoxy buffers the entire document body. This can be potentially dangerous, since a server could just keep sending data indefinitely and wait for your RAM to exhaust -- with nasty consequences. Hence this option.

When a document buffer size reaches the buffer-limit, it is flushed to the client unfiltered and no further attempt to filter the rest of the document is made. Remember that there may be multiple threads running, which might require up to buffer-limit Kbytes each, unless you have enabled "single-threaded" above.

To which parent HTTP proxy specific requests should be routed.

target_pattern http_parent[:port]

where target_pattern is a URL pattern that specifies to which requests (i.e. URLs) this forward rule shall apply. Use / to denote "all URLs". http_parent[:port] is the DNS name or IP address of the parent HTTP proxy through which the requests should be forwarded, optionally followed by its listening port (default: 8080). Use a single dot (.) to denote "no forwarding".

Unset

Don't use parent HTTP proxies.

If http_parent is ".", then requests are not forwarded to another HTTP proxy but are made directly to the web servers.

Multiple lines are OK, they are checked in sequence, and the last match wins.

Everything goes to an example anonymizing proxy, except SSL on port 443 (which it doesn't handle):

forward / anon-proxy.example.org:8080 forward :443 .

Everything goes to our example ISP's caching proxy, except for requests to that ISP's sites:

forward / caching-proxy.example-isp.net:8000 forward .example-isp.net .

Through which SOCKS proxy (and to which parent HTTP proxy) specific requests should be routed.

target_pattern socks_proxy[:port] http_parent[:port]

where target_pattern is a URL pattern that specifies to which requests (i.e. URLs) this forward rule shall apply. Use / to denote "all URLs". http_parent and socks_proxy are IP addresses in dotted decimal notation or valid DNS names (http_parent may be "." to denote "no HTTP forwarding"), and the optional port parameters are TCP ports, i.e. integer values from 1 to 64535

Unset

Don't use SOCKS proxies.

Multiple lines are OK, they are checked in sequence, and the last match wins.

The difference between forward-socks4 and forward-socks4a is that in the SOCKS 4A protocol, the DNS resolution of the target hostname happens on the SOCKS server, while in SOCKS 4 it happens locally.

If http_parent is ".", then requests are not forwarded to another HTTP proxy but are made (HTTP-wise) directly to the web servers, albeit through a SOCKS proxy.

From the company example.com, direct connections are made to all "internal" domains, but everything outbound goes through their ISP's proxy by way of example.com's corporate SOCKS 4A gateway to the Internet.

forward-socks4a / socks-gw.example.com:1080 www-cache.example-isp.net:8080 forward .example.com .

A rule that uses a SOCKS 4 gateway for all destinations but no HTTP parent looks like this:

forward-socks4 / socks-gw.example.com:1080 .