I've just restart my work on my boincs stats mirror (sharing the stats files my site downloads from the various boinc projects). I've put together some zip files by project and update_time at boincs com and it occured to me, that I should ask--what do you need? how would you like the files to be packaged?
I anticipate a stats site would check the update time from a project's tables.xml, check to see if it's been mirrored and if so, download from the mirror. This should reduce the load on the project servers and for many projects decrease the download time for you.
Since I'm working in a vacuum, I just zipped the files from wget named in reverse host order and with the upate time (unix stamp). I have yet to provide the xml? output of the project details and link to the latest archive. That's why I'm checking in here for some advice, suggestions, and hopefully to get a 'user' of the mirror.
Data validation is also on my mind...like tables counts of entries in the xml files--do they agree, etc. As of now it's just doing an archive test of the gz files.
[Update] Ideas for 'proxy' like server--discussion at my place http://www.technutopia.com/forum/showthread.php?t=6737
-- Your participation and assistance would be appreciated--thanks.