Last modified: 2014-09-24 00:05:53 UTC
Created attachment 6722 [details] SquidUpdate, Title, DynamicPageList This code aims to fix bug 20818 squid caching issues. First, SquidUpdate::newFromTitles() is changed to be based on getSquidURLs() instead of generating the url by itself. This probably fixes caching problems on wikis with variants. Second, TitleSquidURLs hook is added to allow extensions to add extra article-related urls which need purging. Third, DynamicPageList is modified to use the new hook to purge the ?curid urls it generates.
Can you please revise the patch, as bug 20818 has been resolved? (or do i have the relations the wrong way around?)
Bug 20818 was fixed by not using curids for Google. In strict sense, curid urls should always be purged. Here it was just doing it when DPL is active, to decrease load. If DPL is handling out curid links, those shouldn't lead to old pages. Perhaps there should be a DPL option to choose if allowing curid urls or not?
Note: There is no longer any need for DPL's to output curid's whatsoever. The only reason it was there in the first place was to have a number in the urls. Since that can just as easily be served as a dummy parameter (which it is currently), DPL really does not need this anymore.
Right, but we may want (or not) to still fix caching problem with curid urls. And there is also the issue of variant wikis.
Platonides, marking your patch as obsolete since it doesn't apply cleanly to trunk anymore. I'm cc'ing Gabriel Wicke in case he wants to consider this issue and your approach in the parser rewrite.
Comment on attachment 6722 [details] SquidUpdate, Title, DynamicPageList This patch no longer applies to trunk per Rusty Burchfield's automated testing https://docs.google.com/spreadsheet/ccc?key=0Ah_71HHl7qa7dGtvSms3TGpHQU9NU2Y1VmNzUEUteWc .
(In reply to comment #5) > Platonides, marking your patch as obsolete since it doesn't apply cleanly to > trunk anymore. I'm cc'ing Gabriel Wicke in case he wants to consider this > issue and your approach in the parser rewrite. This isn't really a parser issue.