The format of filters for the URLBlacklist and URLWhitelist policies, as of Chrome 15, is:
- Scheme can be http, https or ftp. This field is optional, and must be followed by '://'.
- An optional '.' (dot) can prefix the host field to disable subdomain matching, see below for details.
- The host field is required, and is a valid hostname or an IP address. It can also take the special '*' value, see below for details.
- An optional port can come after the host. It must be a valid port value from 1 to 65535.
- An optional path can come at the end. Any string can be used here.
The format is very similar to the URL format, with some exceptions:
- user:pass fields can be included but will be ignored (e.g. http://user:firstname.lastname@example.org/pub/bigfile.iso).
- The host can be '*'. It can also have a '.' as a prefix.
- URL parameters will be ignored.
The filter selected for a URL is the most specific match found:
- First, the filters with the longest host match will be selected;
- Among these, filters with a non-matching scheme or port are discarded;
- Among these, the filter with the longest matching path is selected;
- If no valid filter is left at step 3, the host is reduced by removing the left-most subdomain, and trying again from step 1;
- If a filter is available at step 3, its decision (block or allow) is enforced. If no filter ever matches, the default is to allow the request.
The special '*' host will be the last searched, and matches all hosts. When both a blacklist and whitelist filter apply at step 3 (with the same path length), the whitelist filter takes precedence. If a filter has a '.' (dot) prefixing the host, only exact host matches will be filtered:
- "example.com" matches "example.com", "www.example.com" and "sub.www.example.com";
- ".www.example.com" only matches exactly "www.example.com".
Example of searching for a match for "http://mail.example.com/mail/inbox":
- First find filters for "mail.example.com", and go to step 2. If that fails, then try again with "example.com", "com" and finally "".
- Among the current filters, remove those that have an HTTPS or FTP scheme;
- Among the current filters, remove those that have an exact port number and it not 80;
- Among the current filters, remove those that don't have "/mail/inbox" as a prefix of the path;
- Pick the filter with the longest path prefix, and apply it. If no such filter exists, go back to step 1 and try the next subdomain.
- "example.com" blocks all requests to that domain and any subdomain;
- "http://example.com" blocks all HTTP requests to that domain and any subdomain; HTTPS and FTP requests are still allowed;
- "https://*" blocks all HTTPS requests to any domain;
- "mail.example.com" blocks this domain but not "www.example.com" nor "example.com";
- ".example.com" blocks exactly "example.com", and won't block subdomains;
- "*" blocks all requests; only whitelisted URLs will be allowed;
- "*:8080" blocks all requests to port 8080;
- "example.com/stuff" blocks all requests to any subdomain of "example.com" that have "/stuff" as a prefix of the path;
- "192.168.1.2" blocks requests to this exact IP address.
Example: allowing only a small set of sites:
- Block "*"
- Allow selected sites: "mail.example.com", "wikipedia.org", "google.com"
Example: block all access to a domain, except to the mail server using HTTPS and to the main page:
- Block "example.com"
- Allow "https://mail.example.com"
- Allow ".example.com", and maybe ".www.example.com"