al_9x wrote:
Fx 3.6.16 & !@* & hotmail - about:blank is dumped for the 3rd frame (HistoryFrame.aspx), I am assuming this should not be happening, is that right?
Wrong. Most frames are initially accessible and can be scripted while still hosting "about:blank" (it gets replaced as soon as you actually load something there or use document.open()/write()).
Therefore it's useful for surrogates to run there (the LiveConnect interception pseudo-surrogate, for instance, MUST run on such frames in order to prevent the parent from instantiating Java by referencing the frame's content DOM).
al_9x wrote:
Fx 3.6.16 & @* & hotmail - still a total of 9 invocations on 5 frames, 6 about:blanks. Again, I am assuming there should be only 5 and with proper urls, is that not correct?
Not sure. Are there HTTP redirection?
al_9x wrote:
Fx 4.0.0 & @* & hotmail - still running on just the first three frames
It might depend on parser differences - Fx 4 uses a brand new parser, and frame loading was one of the less specified thing in "legacy" HTML.
al_9x wrote:
Are you not seeing the above?
I'm not checking hotmail (too many variables), I'm using a contrived test case with top document, "normal" frames and scripted frames, and they work as expected both on 3.6.x and 4.0.
On a side note, Fx 4.0 is my reference ATM, since the callback which I depend on to inject the scripts is officially sanctioned, used by Jetpacks, and very likely does "the right thing".
Differences in Fx 3.6 may or may not be bugs because there's no official callback and I use fragile hacks, especially for frames (still they might be due to parser differences, see above), but in my contrived tests there's no significant difference and both the browsers seem to apply surrogates when needed.
al_9x wrote:
I'll do another perf test after the regressions are fixed, should the perf stay the same as that first good rc
OK, thanks. Something I didn't ask previously: did you actually test with cnn.com whitelisted or not? If not, the perf difference is unlikely to depend on surrogate, since ! surrogates are just for a couple of image hosting services, by default.
al_9x wrote:
when you post "try latest build" could you please include the build number, there are times you need to go back through the thread and pick out a specific build, addressing what's being discussed
OK.
Mozilla/5.0 (Windows; U; Windows NT 5.2; en-US; rv:1.9.2.16) Gecko/20110319 Firefox/3.6.16
