Michael Rosenthal schrieb:
I suggest keep the bug on Wikimedia's servers and using a tool which relies on SQL databases. These could be shared with the toolserver where the "official" version of the analysis tool runs and users are enabled to run their own queries (so taking a tool with a good database structure would be nice). With that the toolserver users could set up their own cool tools on that data.
Well, the original problem is that wikipedia has so many page views, writing each to a database will simply melt that database. we are talking about 50000 hits per second. this is of course also true for the toolserver.
I was thinking about a solution that uses sampling, or would only be applied to specific pages or small projects. We hat something similar for the old wikicharts.
-- daniel