Last modified: 2012-04-24 10:19:05 UTC
Currently we got this: 21:03 < rc2012> [[Translations:Submissions/22/es]] !N http://wikimania2012.wikimedia.org/w/index.php?oldid=19055&rcid=20442 * Igna * (+218) trad 21:03 < rc2012> [[Submissions/es]] http://wikimania2012.wikimedia.org/w/index.php?diff=19056&oldid=19054 * Igna * (+14) trad 21:05 < rc2012> [[Translations:Submissions/24/es]] !N http://wikimania2012.wikimedia.org/w/index.php?oldid=19057&rcid=20444 * Igna * (+151) trad 21:05 < rc2012> [[Submissions/es]] http://wikimania2012.wikimedia.org/w/index.php?diff=19058&oldid=19056 * Igna * (+24) trad 21:07 < rc2012> [[Translations:Submissions/26/es]] !N http://wikimania2012.wikimedia.org/w/index.php?oldid=19059&rcid=20446 * Igna * (+138) trad 21:07 < rc2012> [[Submissions/es]] http://wikimania2012.wikimedia.org/w/index.php?diff=19060&oldid=19058 * Igna * (+15) trad 21:07 < rc2012> [[Translations:Submissions/26/es]] http://wikimania2012.wikimedia.org/w/index.php?diff=19061&oldid=19059 * Igna * (+18) 21:07 < rc2012> [[Submissions/es]] http://wikimania2012.wikimedia.org/w/index.php?diff=19062&oldid=19060 * Igna * (+18) Not sure what those "Translations:Submissions/26/es" things are. Those always come in pairs, are they duplicates? Besides, I think it would be nice to have a special flag on translations, like "N", or "B" - maybe "T"?
Since IRC output has to be considered as a stable API there isn't anything we can do here.
WONTFIXing per comment #1
Come on, let's not give up so easily. Special:Recentchanges knows somehow how to filter out translations First, I am not sure (but I don't know the code) why we have two log entries instead of one (may be one does not need one of them)? Second, maybe some specific action can be logged with rc_action_type instead of normal edit summary? FlaggedRevs is doing this (although reviews are not edits, translations are) and this is perfect for IRC purposes. Third, there is a checkuser logging, which does not work very well because of FauxRequest and bug 34838. Updating description to indicate a slightly wider problem.
Both edits are real. There is nothing I can do about it.
They are not duplicates Marcin. It is carrying out two processes - logging what it's doing, and posting it. There's two 26's because the bot seems to have done two passes - it didn't have a page in translation to link to, so on the second pass it linked to the page in English and piped into Spanish. They are clearly two edits. Could you explain how the other two problems it's causing are related here. I've been following the email thread, so I know 34838 is where localhost is appearing in the records.
(In reply to comment #5) > They are not duplicates Marcin. It is carrying out two processes - logging what > it's doing, and posting it. There's two 26's because the bot seems to have done > two passes - it didn't have a page in translation to link to, so on the second > pass it linked to the page in English and piped into Spanish. They are clearly > two edits. I don't have a problem with two 21:07 < rc2012> [[Translations:Submissions/26/es]] !N http://wikimania2012.wikimedia.org/w/index.php?oldid=19059&rcid=20446 and 21:07 < rc2012> [[Translations:Submissions/26/es]] http://wikimania2012.wikimedia.org/w/index.php?diff=19061&oldid=19059 * Igna * (+18) I find this strange: 21:05 < rc2012> [[Submissions/es]] http://wikimania2012.wikimedia.org/w/index.php?diff=19058&oldid=19056 * Igna * (+24) trad is always followed by: 21:07 < rc2012> [[Translations:Submissions/26/es]] !N http://wikimania2012.wikimedia.org/w/index.php?oldid=19059&rcid=20446 Those always come in pairs. For some reason, actually two items in the revision table are created. But did the user did ONE edit here or two? > Could you explain how the other two problems it's causing are related here. > I've been following the email thread, so I know 34838 is where localhost is > appearing in the records. Second problem: "If the second revision logged (Translations:Submissions/26/es) is an internal request/action to create another revision and not the user's action, it should be something other as the edit - some internal logged action". Third problem: "Those internal actions (obviously they are not edits) also cause bug 34838 - they don't come from the user's browser but are internally generated on his/her behalf". We should decide: * Either those TWO revisions are REAL user edits (user clicked on something twice and saved two changes) - in this cases BOTH should be logged as proper edits with user's IP address and the user agent - this is currently not the case because of bug 34838 (FauxRequest user) OR * The second revision is somehow automatically generated by the software for some reasons and is *not* executed by the user, only on its behalf. In this case it should not be then logged as an edit but as an internal action; again with proper user data (bug 34828). I didn't have time to dig into the architecture of the code and find out why those two editions are actually necessary; *IF* they are not necessary, one revision should be recorded only; *IF* they are necessary, the above decision needs to be made. FauxRequest (bug 34828) should be fixed independently of this issue.
Er, I meand bug 34838 all the time above, two last links is incorrect.
Those are two real pages (you can go and open them and see they have different history). See https://www.mediawiki.org/wiki/Help:Extension:Translate/Page_translation_administration#Anatomy_of_a_translatable_page The whole point of them is to keep history, so it wouldn't make sense to hide the real user.
