Yep, but bytecounts as an approximate for information density or content size are themselves not terribly useful (mobile web or desktop, two different bytecounts, same content). The goal with this is more, I think, to enable a reference point to work out "okay, what version of the article were these people prrrobably looking at, and what did it look like?"
On 20 March 2015 at 00:18, Gergo Tisza gtisza@wikimedia.org wrote:
On Mon, Mar 16, 2015 at 3:14 PM, Oliver Keyes okeyes@wikimedia.org wrote:
Kevin: I'm not sure what value there'd be. I mean, there's page-size, maybe? But pageID gives us that (or should).
Time-traveling with MediaWiki is very hard. Calculating the length of wikitext for a given pageID at a given time is cumbersome (instead of simple text processing, you are now dealing with DB queries, need to set up a local DB mirror etc). Finding out what title it had at the time is prohibitively hard (you have to parse semi-structured objects which are serialized into strings in the log table and follow the chain of renames). Finding out the byte size of the rendered HTML is practically impossible (templates and interface messages change; flagged revisions/pending changes might result in older versions of articles being shown). If you omit the bytecounts, there is no way people will be able to reconstruct them from the logs.
Not saying that's a problem - I personally don't see much use for them. Just don't expect pageID to be very useful for "normalizing" logs.
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics