[go: up one dir, main page]

About this service

In order to check the validity of the technical reports that W3C publishes, the Systems Team has developed a link checker.

A first version was developed in by Renaud Bruyeron. Since it was lacking some functionalities, Hugo Haas rewrote it more or less from scratch in . It has been improved by Ville Skyttä and many other volunteers since.

The source code is available publicly under the W3C IPR software notice from CPAN (released versions) and a github repository (development and archived release versions).

What it does

The link checker reads an HTML or XHTML document or a CSS style sheet and extracts a list of anchors and links.

It checks that no anchor is defined twice.

It then checks that all the links are dereferenceable, including the fragments. It warns about HTTP redirects, including directory redirects.

It can check recursively a part of a Web site.

There is a command line version and a CGI version. They both support HTTP basic authentication. This is achieved in the CGI version by passing through the authorization information from the user browser to the site tested.

Use it online

There is an online version of the link checker.

In the online version (and in general, when run as a CGI script), the number of documents that can be checked recursively is limited.

Both the command line version and the online one sleep at least one second between requests to each server to avoid abuses and target server congestion.

Access keys

The following access keys are implemented throughout the site in an attempt to help users using screen readers.

  1. Home: access key 1 leads back to the service's home page.
  2. Downloads: access key 2 leads to downloads.
  3. Documentation: access key 3 leads to the documentation index for the service.
  4. Feedback: access key 4 leads to the feedback instructions.

Install it locally

The link checker is written in Perl. It is packaged as a standard CPAN distribution, and depends on a few other modules which are also available from CPAN.

Install with the CPAN utility

If you system has a working installation of Perl, you should be able to install the link checker and its dependencies with a single line from the commandline shell:

sudo perl -MCPAN -e 'install W3C::LinkChecker' (use without the sudo command if installing from an administrator account).

If this is the first time you use the CPAN utility, you may have to answer a few setup questions before the tool downloads, builds and installs the link checker.

Install by hand

If for any reason the technique described above is not working or if you prefer installing each package by hand, follow the instructions below:

  1. Install Perl, version 5.8 or newer.
  2. You will need the following CPAN distributions, as well as the distributions they possibly depend on. Depending on your Perl version, you might already have some of these installed. Also, the latest versions of these may require a recent version of Perl. As long as the minimum version requirement(s) below are satisfied, everything should be fine. The latest version should not be needed, just get an older version that works with your Perl. For an introduction to installing Perl modules, see The CPAN FAQ.
  3. Optionally install the link checker configuration file, etc/checklink.conf contained in the link checker distribution package into /etc/w3c/checklink.conf or set the W3C_CHECKLINK_CFG environment variable to the location where you installed it.
  4. Optionally, install the checklink script into a location in your web server which allows execution of CGI scripts (typically a directory named cgi-bin somewhere below your web server's root directory).
  5. See also the README and INSTALL file(s) included in the above distributions.

Running checklink --help shows how to use the command line version. The distribution package also includes more extensive POD documentation, use perldoc checklink (or man checklink on Unixish systems) to view it.

SSL/TLSv1 support for https in the link checker needs support for it in libwww-perl; see README.SSL in the libwww-perl distribution for more information.

In online mode, link checker's output should not be buffered to avoid browser timeouts. The link checker itself does not buffer its output, but in some cases output buffering needs to be explicitly disabled for it in the web server running it. One such case is Apache's mod_deflate compression module which as a side effect results in output buffering; one way to disable it for the link checker (while leaving it enabled for other resources if configured so elsewhere) is to add the following section to an appropriate place in the Apache configuration (assuming the link checker script's filename is checklink): 

        <Files checklink>
        SetEnv no-gzip
        </Files>

If you want to enable the authentication capabilities with Apache, have a look at Steven Drake's hack.

The link checker honors proxy settings from the scheme_proxy environment variables. See LWP(3) and LWP::UserAgent(3)'s env_proxy method for more information.

Some environment variables affect the way how the link checker uses FTP. In particular, passive mode is the default. See Net::FTP(3) for more information.

There are multiple alternatives for configuring the default NNTP server for use with news: URIs without explicit hostnames, see Net::NNTP(3) for more information.

Robots exclusion

The link checker honors robots exclusion rules. To place rules specific to the W3C Link Checker in /robots.txt files, sites can use the W3C-checklink user agent string. For example, to allow the link checker to access all documents on a server and to disallow all other robots, one could use the following: 

          User-Agent: *
          Disallow: /
          User-Agent: W3C-checklink
          Disallow:

Robots exclusion support in the link checker is based on the LWP::RobotUA Perl module. It currently supports the "original 1994 version" of the standard. The robots META tag, ie. <meta name="robots" content="...">, is not supported. Other than that, the link checker's implementation goes all the way in trying to honor robots exclusion rules; if a /robots.txt disallows it, not even the first document submitted as the root for a link checker run is fetched.

Note that /robots.txt rules affect only user agents that honor it; it is not a generic method for access control.

Comments, suggestions and bugs

The current version has proven to be stable. It could however be improved, see the list of open enhancement ideas and bugs for details.