- From: Henri Sivonen <hsivonen@iki.fi>
- Date: Tue, 24 Mar 2009 14:07:47 +0200
- To: Anne van Kesteren <annevk@opera.com>
- Cc: "Henry S. Thompson" <ht@inf.ed.ac.uk>, "Ian Hickson" <ian@hixie.ch>, "HTML WG" <public-html@w3.org>
On Mar 23, 2009, at 20:17, Anne van Kesteren wrote: > On Mon, 23 Mar 2009 18:53:39 +0100, Henry S. Thompson > <ht@inf.ed.ac.uk> wrote: >> Ian wrote: >>> My understanding is that doing this would introduce an unacceptable >>> performance penalty for implementations. >> >> So, I'm perplexed. >> >> All the browsers I'm familiar with (I haven't installed Chrome yet) >> support XML+XSLT via the xml-stylesheet processing instruction. >> >> So they all have case-preserving tokenizers whose performance is >> acceptable for the people who serve and read XML+XSLT. >> >> Who then has ruled that a case-preserving tokenizer "imposes an >> unacceptable performance penalty", and on the basis of what evidence? > > Henri elaborates that in the very next e-mail of that thread (and > has elaborated on it in the past as well): > > http://lists.w3.org/Archives/Public/public-html/2009Mar/0240.html > > (Second instance of "Indeed" has the details.) I should add that the pre-interned well-known token objects also carry a carefully chosen integer that the tree builder switches on to decide which behavior case to run for a start or end tag token. Having integers like this on the tokens was one reason for the pre-interned token object design. See for example: http://hg.mozilla.org/users/mrbkap_mozilla.com/html5parsing/file/6611e8bac15f/content/html/parser/src/nsHtml5ElementName.cpp#l251 http://hg.mozilla.org/users/mrbkap_mozilla.com/html5parsing/file/6611e8bac15f/content/html/parser/src/nsHtml5TreeBuilder.cpp#l893 In the future, I intend to[1] add DOM node creation function pointers to the token objects to avoid a later lookup when the right constructor needs to be called to create right class instances for nodes that have element-specific subclasses in the DOM implementation. [1] https://bugzilla.mozilla.org/show_bug.cgi?id=483155 -- Henri Sivonen hsivonen@iki.fi http://hsivonen.iki.fi/
Received on Tuesday, 24 March 2009 12:08:29 UTC