[Webkit-unassigned] [Bug 27165] Connections-per-host should be tracked closer to the ResourceHandle level, not the Cache

Mon Nov 8 10:44:00 PST 2010

https://bugs.webkit.org/show_bug.cgi?id=27165

--- Comment #42 from Nate Chapin <japhet at chromium.org>  2010-11-08 10:44:00 PST ---
(In reply to comment #41)
> (From update of attachment 72883 [details])
> I think that this patch is flawed in that it changes load scheduling instead of just the promised refactoring. It's the fact that XMLHttpRequest loads do not start right away - even when all loader slots are empty - that causes the test regression.
> 
> That said, this patch still seems like an improvement all things considered. Given Nate's good track record of thoroughness in particular, I think this is OK to land it and investigate XMLHttpRequest scheduling changes in a follow-up bug.

I just experimented with bumping XHRs from Low priority to Medium (which guarantees we'll try to start them immediately), and that causes this test to pass.  If it's ok with you, I think I'll add changing the ResourceLoaderScheduler::Low to Medium in DocumentThreadableLoader.cpp to the list of changes before submitting.

Sorry for not catching this sooner, I really thought I had tried fixing the problem by just increasing the priority of XHRs.  Note, however, that this test could theoretically become flaky if we do hit the number of requests where we start throttling.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.