- From: Maciej Stachowiak <mjs@apple.com>
- Date: Sun, 14 Nov 2010 08:20:32 -0800
- To: Julian Reschke <julian.reschke@gmx.de>
- Cc: Jonas Sicking <jonas@sicking.cc>, Ian Hickson <ian@hixie.ch>, "public-html@w3.org" <public-html@w3.org>
On Nov 14, 2010, at 3:43 AM, Julian Reschke wrote: > On 14.11.2010 05:15, Jonas Sicking wrote: >> ... >> The best solution to a whole group of problems here is IMHO to define >> that<meta http-equiv> has no relation to HTTP headers at all. Any and >> all similarities with http and http headers is a historical artifact. >> ... > > If that's what we think, we should clearly say that. This particular algorithm is even more limited in scope than most http-equiv values, since (according to the backwards cross-reference tool) it is only used by the parser to determine the document's character set encoding. That code is historically quite different from actual Content-Type parsing code. > > That would mean clarifying that the section is *only* about meta/@http-equiv, and clearly state that *because* it's not about the HTTP header field the parsing rules can vary. If such a statement was added, would you consider that sufficient to resolve this issue and ISSUE-126 by amicable resolution? > > That being said: even if we do that it would be good to reduce *unnecessary* deviations. For instance, it's totally not clear why "foocharset" is parsed as "charset", while "charsetfoo" is not (<http://www.w3.org/Bugs/Public/show_bug.cgi?id=9628#c3>). It seems like that is a separate concern from the two issues currently under discussion. That being said, I believe older (pre-HTML5 parser) browsers generally work that way. When detecting the encoding, once they see "<meta", pre-HTML5 browsers just scan forward to find "charset=" before hitting ">". That's somewhat oversimplified, but a decent first-order approzimation. From that model, you can see why foocharset would be detected and charsetfoo would not. This same looseness is what makes HTML5's simplified charset syntax (<meta charset=utf8>) work in current browsers. If any case, if we want to take up this detail further, it should be via a separate bug/issue. Regards, Maciej
Received on Sunday, 14 November 2010 16:21:07 UTC