Apache Module mod_filter

Smart Filtering

In the traditional filtering model, filters are inserted unconditionally using AddOutputFilter and family. Each filter then needs to determine whether to run, and there is little flexibility available for server admins to allow the chain to be configured dynamically.

mod_filter by contrast gives server administrators a great deal of flexibility in configuring the filter chain. In fact, filters can be inserted based on complex boolean expressions This generalises the limited flexibility offered by AddOutputFilterByType.

Filter Declarations, Providers and Chains

Figure 1: The traditional filter model

In the traditional model, output filters are a simple chain from the content generator (handler) to the client. This works well provided the filter chain can be correctly configured, but presents problems when the filters need to be configured dynamically based on the outcome of the handler.

Figure 2: The mod_filter model

mod_filter works by introducing indirection into the filter chain. Instead of inserting filters in the chain, we insert a filter harness which in turn dispatches conditionally to a filter provider. Any content filter may be used as a provider to mod_filter; no change to existing filter modules is required (although it may be possible to simplify them). There can be multiple providers for one filter, but no more than one provider will run for any single request.

A filter chain comprises any number of instances of the filter harness, each of which may have any number of providers. A special case is that of a single provider with unconditional dispatch: this is equivalent to inserting the provider filter directly into the chain.

Configuring the Chain

There are three stages to configuring a filter chain with mod_filter. For details of the directives, see below.

Declare Filters

The FilterDeclare directive declares a filter, assigning it a name and filter type. Required only if the filter is not the default type AP_FTYPE_RESOURCE.

Register Providers

The FilterProvider directive registers a provider with a filter. The filter may have been declared with FilterDeclare; if not, FilterProvider will implicitly declare it with the default type AP_FTYPE_RESOURCE. The provider must have been registered with ap_register_output_filter by some module. The final argument to FilterProvider is an expression: the provider will be selected to run for a request if and only if the expression evaluates to true. The expression may evaluate HTTP request or response headers, environment variables, or the Handler used by this request. Unlike earlier versions, mod_filter now supports complex expressions involving multiple criteria with AND / OR logic (&& / ||) and brackets. The details of the expression syntax are described in the ap_expr documentation.

Configure the Chain

The above directives build components of a smart filter chain, but do not configure it to run. The FilterChain directive builds a filter chain from smart filters declared, offering the flexibility to insert filters at the beginning or end of the chain, remove a filter, or clear the chain.

Filtering and Response Status

mod_filter normally only runs filters on responses with HTTP status 200 (OK). If you want to filter documents with other response statuses, you can set the filter-errordocs environment variable, and it will work on all responses regardless of status. To refine this further, you can use expression conditions with FilterProvider.

Upgrading from Apache HTTP Server 2.2 Configuration

The FilterProvider directive has changed from httpd 2.2: the match and dispatch arguments are replaced with a single but more versatile expression. In general, you can convert a match/dispatch pair to the two sides of an expression, using something like:

The Request headers, Response headers and Environment variables are now interpreted from syntax %{req:foo}, %{resp:foo} and %{env:foo} respectively. The variables %{HANDLER} and %{CONTENT_TYPE} are also supported.

Note that the match no longer support substring matches. They can be replaced by regular expression matches.

Examples

Server side Includes (SSI)

A simple case of replacing AddOutputFilterByType

FilterDeclare SSI
FilterProvider SSI INCLUDES "%{CONTENT_TYPE} =~ m|^text/html|"
FilterChain SSI

Server side Includes (SSI)

The same as the above but dispatching on handler (classic SSI behaviour; .shtml files get processed).

FilterProvider SSI INCLUDES "%{HANDLER} = 'server-parsed'"
FilterChain SSI

Emulating mod_gzip with mod_deflate

Insert INFLATE filter only if "gzip" is NOT in the Accept-Encoding header. This filter runs with ftype CONTENT_SET.

FilterDeclare gzip CONTENT_SET
FilterProvider gzip inflate "%{req:Accept-Encoding} !~ /gzip/"
FilterChain gzip

Image Downsampling

Suppose we want to downsample all web images, and have filters for GIF, JPEG and PNG.

FilterProvider unpack jpeg_unpack "%{CONTENT_TYPE} = 'image/jpeg'"
FilterProvider unpack gif_unpack  "%{CONTENT_TYPE} = 'image/gif'"
FilterProvider unpack png_unpack  "%{CONTENT_TYPE} = 'image/png'"

FilterProvider downsample downsample_filter "%{CONTENT_TYPE} = m|^image/(jpeg|gif|png)|"
FilterProtocol downsample "change=yes"

FilterProvider repack jpeg_pack "%{CONTENT_TYPE} = 'image/jpeg'"
FilterProvider repack gif_pack  "%{CONTENT_TYPE} = 'image/gif'"
FilterProvider repack png_pack  "%{CONTENT_TYPE} = 'image/png'"
<Location "/image-filter">
    FilterChain unpack downsample repack
</Location>

Protocol Handling

Historically, each filter is responsible for ensuring that whatever changes it makes are correctly represented in the HTTP response headers, and that it does not run when it would make an illegal change. This imposes a burden on filter authors to re-implement some common functionality in every filter:

• Many filters will change the content, invalidating existing content tags, checksums, hashes, and lengths.
• Filters that require an entire, unbroken response in input need to ensure they don't get byteranges from a backend.
• Filters that transform output in a filter need to ensure they don't violate a Cache-Control: no-transform header from the backend.
• Filters may make responses uncacheable.

mod_filter aims to offer generic handling of these details of filter implementation, reducing the complexity required of content filter modules. This is work-in-progress; the FilterProtocol implements some of this functionality for back-compatibility with Apache 2.0 modules. For httpd 2.1 and later, the ap_register_output_filter_protocol and ap_filter_protocol API enables filter modules to declare their own behaviour.

At the same time, mod_filter should not interfere with a filter that wants to handle all aspects of the protocol. By default (i.e. in the absence of any FilterProtocol directives), mod_filter will leave the headers untouched.

At the time of writing, this feature is largely untested, as modules in common use are designed to work with 2.0. Modules using it should test it carefully.