Last Comment Bug 602518 - Need Hint to re-enable pipeline on XHR
: Need Hint to re-enable pipeline on XHR
Status: RESOLVED WONTFIX
:
Product: Core
Classification: Components
Component: Networking: HTTP (show other bugs)
: Trunk
: x86_64 Linux
: -- normal with 2 votes (vote)
: ---
Assigned To: Patrick McManus [:mcmanus]
:
: Patrick McManus [:mcmanus]
Mentors:
Depends on: 599164
Blocks: 603503 603505 pipelining-review
  Show dependency treegraph
 
Reported: 2010-10-07 08:22 PDT by Patrick McManus [:mcmanus]
Modified: 2015-12-16 13:13 PST (History)
10 users (show)
See Also:
Crash Signature:
(edit)
QA Whiteboard:
Iteration: ---
Points: ---
Has Regression Range: ---
Has STR: ---


Attachments
Allow X-UA-Response-Speed: fast to bypass SLOW heuristic (2.06 KB, patch)
2010-10-07 08:32 PDT, Patrick McManus [:mcmanus]
no flags Details | Diff | Splinter Review
Allow X-UA-Response-Speed: fast to bypass SLOW heuristic (2.25 KB, patch)
2010-12-03 15:25 PST, Patrick McManus [:mcmanus]
no flags Details | Diff | Splinter Review
Allow X-UA-Response-Speed: fast to bypass SLOW heuristic (2.25 KB, patch)
2011-06-26 22:07 PDT, Patrick McManus [:mcmanus]
no flags Details | Diff | Splinter Review

Description Patrick McManus [:mcmanus] 2010-10-07 08:22:41 PDT
The patch in bug 599164 disables pipelines for XHR transactions.. Long polls are one of the major pipeline interaction problems, so that is sensible.

OTOH there certainly are things done from XHR that fit the pipeline pattern well. So we should provide a hint that JS authors can use to tell the HTTP engine that all the resources are "fast" in a pipelineable sense.

I think there are a bunch of different ways to do this so I have split this patch out from 599164 so as not to block that work.
Comment 1 Patrick McManus [:mcmanus] 2010-10-07 08:32:18 PDT
Created attachment 481506 [details] [diff] [review]
Allow X-UA-Response-Speed: fast to bypass SLOW heuristic

This is one approach to the problem. Allow XHR (or any channel user) to do this:

httpRequest.setRequestHeader('X-UA-Response-Speed', 'fast')

which bypasses the SLOW indicator that is placed on by default for XHR.

It avoids changing IDLs, prefixes, detection code, etc.. if it is executed in a context that doesn't understand it the result is a harmless nop.
Comment 2 Patrick McManus [:mcmanus] 2010-12-03 15:25:46 PST
Created attachment 495141 [details] [diff] [review]
Allow X-UA-Response-Speed: fast to bypass SLOW heuristic

bitrot and styl updates..
Comment 3 Brian Smith (:briansmith, :bsmith, use NEEDINFO?) 2011-05-25 15:00:58 PDT
IMO, a request header is not great. It makes every request bigger. The disadvantages of an API call mentioned above seem small. An API call as a hint also seems to fit better with HTML rel mechanism that mnot suggested in his pipelining hints draft.
Comment 4 Patrick McManus [:mcmanus] 2011-06-26 22:07:37 PDT
Created attachment 542082 [details] [diff] [review]
Allow X-UA-Response-Speed: fast to bypass SLOW heuristic

update bitrot to reflect larch
Comment 5 Honza Bambas (:mayhemer) 2011-11-03 13:09:49 PDT
Comment on attachment 542082 [details] [diff] [review]
Allow X-UA-Response-Speed: fast to bypass SLOW heuristic

Dropping the review request for now, just because I'm cleaning my review queue, before we get to the pipelining stuff again.  Let's discuss this particular feature more widely first.
Comment 6 Brian Smith (:briansmith, :bsmith, use NEEDINFO?) 2011-12-29 18:08:38 PST
During discussion of how to optimize our addons update check, I suggested that we may be able to add a hint to cause Necko to be optimistic about pipelining for a connection to reduce the number of parallel connections to versioncheck.addons.mozilla.org during update checking. If the addon update checker uses XHR (probably) then having such an option might be part of solving that problem.
Comment 7 Patrick McManus [:mcmanus] 2015-12-16 13:13:21 PST
if anything - we will rip out pipelines

Note You need to log in before you can comment on or make changes to this bug.