- From: Patrick Meenan <pmeenan@webpagetest.org>
- Date: Fri, 5 Dec 2014 17:53:41 -0500
- To: Boris Zbarsky <bzbarsky@mit.edu>
- Cc: "public-web-perf@w3.org" <public-web-perf@w3.org>
Received on Friday, 5 December 2014 22:54:12 UTC
Blink also has a light tree pass that's part of the background parser <https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit/Source/core/html/parser/BackgroundHTMLParser.cpp&sq=package:chromium&l=226> that the preloader runs off of. As long as the optional information was tagged on the elements directly (not applied by styles) it shouldn't be too hard to make blink skip preloading for the whole tree below it. On Fri, Dec 5, 2014 at 5:43 PM, Boris Zbarsky <bzbarsky@mit.edu> wrote: > On 12/5/14, 2:16 PM, Ilya Grigorik wrote: > >> And/or we just need to rethink how the Blink preloader operates. AFAIK, >> FF (and IE, I believe) construct the actual tree as part of the preload >> scan -- FF/IE folks, please correct me if that's not true. >> > > What Firefox does is that we just run the tokenizer and generate tree > construction commands but do not actually perform the tree construction. > So the output here is not a tree, but a list of commands, but a tree > structure can be recovered from it. > > -Boris > >
Received on Friday, 5 December 2014 22:54:12 UTC