Comments on: Meaningful estimates of popularity.

By: Tgr

Tgr — Sun, 25 Nov 2007 20:41:21 +0000

Summing up the time spent reading each page would be a good metric, but it would require advanced javascript trickery to get even remotely accurate numbers unless most visitors view several wikipedia pages in succession (wich is probably not the case).

By: Daniel Kinzler

Daniel Kinzler — Sun, 25 Nov 2007 20:04:32 +0000

Yea, I know how it works, I helped Leon design it :) There’s a new version comming up, cased on the UDP log stream – that would no longer require sampling, but would count all hits accurately, for all wikis. I don’t know if the UDP log packets contain the referer though – would have to ask Mark I guess. Anyway, Leon isn’t online much lately (real life calling, maybe)… so no idea when this will be up.

By: David Gerard

David Gerard — Sun, 25 Nov 2007 16:33:58 +0000

It’s possible WikiCharts could do that. (It works by sampling one hit out of every thousand, so that the flood of data doesn’t knock the toolserver over ;-) Float the idea on wikitech-l?

By: Daniel Kinzler

Daniel Kinzler — Sat, 24 Nov 2007 21:08:34 +0000

Filtering by referer would be interesting: only look at clicks that come from inside wikimedia. This would probably be biased the other way around: it would ignore all single page views comming from google etc. In any case, I think it would be nice to compare the two stats.