XMLTokenizer does not parse incrementally as it should. Currently it builds everything up and then does one big parse in finish(). This is bad for lots of reasons. One of them being that it completely kills performance of large files. Once this is fixed, we'll need to look into running <script> tags as we parse them, instead of all at once at the end.
Created attachment 6451 [details] fixes the xml parser to parse incremental (accounting for the xslt fallout) This patch does a few things: 1. removes (unused) onHold support from Tokenizer 2. makes the xml parser parse incrementally. 3. adds serialization of DTDs for xml documents (I added this in my first attempts at fixing xslt) 4. Fixes xslt support to no longer depend on synchronous parsing.
Created attachment 6479 [details] Fixes incremental xml parsing in all cases (includes test results and ChangeLog)
Comment on attachment 6479 [details] Fixes incremental xml parsing in all cases (includes test results and ChangeLog) I gave Eric a few comments in person, but none that seem to block landing this as-is. r=me (Eliminate XSLStyleSheetImpl::setDocument, write test for bug fix in DocumentImpl::recalcStyleSelector that affects recursive transform with CSS style sheets.)
Mass moving XML DOM bugs to the "DOM" Component.