1459566 - Make "frameAncestors" and "proxyInfo" optional in webRequest details object

Status: RESOLVED WONTFIX

(WebExtensions :: Request Handling, enhancement)

Description

[kernp25]

Maybe this will improve performance?

Because, not all extensions need this information.

Then it would look like:
browser.webRequest.onBeforeRequest.addListener(
  test,
  {urls: ["<all_urls>"]},
  ["frameAncestors"]
);

Comment 1

[kernp25]

What do you think?

Comment 2

[Shane Caraveo (:mixedpuppy)]

Do you have some numbers that show its a perf problem?  The heavy lifting is already done in loadinfo.

Comment 3

[kernp25]

(In reply to Shane Caraveo (:mixedpuppy) from comment #2)
> Do you have some numbers that show its a perf problem?  The heavy lifting is
> already done in loadinfo.

I do not have any numbers :)

It was just an idea how to improve maybe the performance.

Because, i was looking at this code:
https://searchfox.org/mozilla-central/rev/b28b94dc81d60c6d9164315adbd4a5073526d372/toolkit/modules/addons/WebRequestContent.js#157-167
https://searchfox.org/mozilla-central/rev/b28b94dc81d60c6d9164315adbd4a5073526d372/toolkit/components/extensions/webrequest/ChannelWrapper.cpp#582-634
https://searchfox.org/mozilla-central/rev/b28b94dc81d60c6d9164315adbd4a5073526d372/toolkit/components/extensions/webrequest/ChannelWrapper.cpp#811-844

and this code looks really large.

Comment 4

[Shane Caraveo (:mixedpuppy)]

(In reply to kernp25 from comment #3)
> (In reply to Shane Caraveo (:mixedpuppy) from comment #2)
> > Do you have some numbers that show its a perf problem?  The heavy lifting is
> > already done in loadinfo.
> 
> I do not have any numbers :)
> 
> It was just an idea how to improve maybe the performance.
> 
> Because, i was looking at this code:
> https://searchfox.org/mozilla-central/rev/
> b28b94dc81d60c6d9164315adbd4a5073526d372/toolkit/modules/addons/
> WebRequestContent.js#157-167

That will only ever be hit by non-http requests, as such I don't see a performance implication here

> https://searchfox.org/mozilla-central/rev/
> b28b94dc81d60c6d9164315adbd4a5073526d372/toolkit/components/extensions/
> webrequest/ChannelWrapper.cpp#582-634
> https://searchfox.org/mozilla-central/rev/
> b28b94dc81d60c6d9164315adbd4a5073526d372/toolkit/components/extensions/
> webrequest/ChannelWrapper.cpp#811-844
> 
> and this code looks really large.

These should only be hit when you actually access the properties.  Of course, that happens in webrequest.jsm anyway whenever a listener exists.  The values are cached, so the hit is only taken once.  I don't think there is real overhead here, but given metrics showing otherwise we should do something about it.  IIRC there are talos tests that use webrequest listeners, so those could probably be used to test whether there is any significant impact.

Comment 5

[Shane Caraveo (:mixedpuppy)]

Given metrics that illustrate a problem we can revisit.