Repeat View timings

concerning this test:

looking at the repeat view the timings seem a little odd.

static caching has been turned on and so the items on the page other than the html are retrieved from browser cache.

(in serial order - to my understanding)
0.269s to get first byte
1.252s to start rendering
0.747s to document complete
0.747s to fully loaded

The screenshots of the page show that the page is complete at both start rendering and fully loaded.

It looks like there is something fishy with the timings. Can someone offer an explanation/clarification?

The screen shot is actually a forced drawing of the browser window to my own in-memory image buffer so it’s usually the same as what was actually on the screen but in some cases IE hasn’t actually painted the screen yet. The start render timing hooks the actual browser window and watches for the actual drawing messages (which is sometimes done in a lazy fashion if there is a lot of activity going on because it is painted in idle cycles).

Hope that helps (and sorry about the confusion).

I’m still a little confused about the timings.

Should I assume that there is some processing going on that is preventing the page from starting to render until 1.252s?

Does the browser fetch resouces in the same way from the browser cache as it would from an external webserver? (just much fast since no external connection needed)
If so, should the document complete and fully loaded timings not be greater than 1.252s? (this assumes that there is an issue with JS execution time and that this would block fetching of the other resources until processing has finished.)

The browser is fetching all of the resources from the cache (the browser itself actually blindly requests all of the resources and then WinInet under the covers serves some from the local cache). My educated guess would be that the requests being served from the cache are at a fast enough rate that the browser doesn’t get enough idle time to actually paint the screen until it has finished all of the processing (though the difference is quite big and I’m not sure what IE would be doing after the Doc Complete that would keep it busy for that long).

I’ve got a fair bit on my plate right now but if I get some spare cycles I’ll try loading the site with a debug build of pagetest which dumps a lot more information about what the browser is doing.