ao2.it
Antonio Ospite, theorist attacks

Harald started migrating the gitweb installation on openezx.org over to cgit, because gitweb was eating up too much resources when stressed by web crawlers; there were still some details to take care of and I put some time into that.

We wanted to:

• Keep the links created by the git-notify mails working;
• Keep most of the URLs indexed by search engines working.

URLs rewriting was the most obvious short term solution.

I found these rules from ClearChain for apache which were used for a similar migration at freedesktop.org, they were still not perfect for us:

• there were some typos most likely due to copy and paste of similar rules;
• the regular expression used for project names ([^.]+)(\.git)* was not working for us because we have project names containing dots (like motorola-2.4);
• the rules were not handling the case where clean URLs were enabled in the gitweb installation.

So I tweaked the rules a little bit and I am now putting them into a gitweb_cgit_migration git repository for other to play with, they can be improved in several ways:

• the project name regex is now assuming that a project name ends in .git which was OK in our case but it might be too strict for other scenarios;
• rules with the same destination address can be merged;
• regex for project names and for commit hashes can be made stricter;
• some non essential pattern groups in the source URL patterns can be removed, but the parameters in the destination URLs would need to be adjusted as well;
• ordering of rules can be improved, right now there might still be some addresses which we are not handling correctly;
• redirects can be made permanent.

However I am not putting more effort in that for now as the result looks acceptable for OpenEZX; anyhow patches are greatly welcome, as always. :)