- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Fri, 04 May 2007 13:48:13 +0200
- To: Anne van Kesteren <annevk@opera.com>
- CC: HTML WG <public-html@w3.org>
Anne van Kesteren wrote: > > On Fri, 04 May 2007 13:09:30 +0200, Julian Reschke > <julian.reschke@gmx.de> wrote: >> But what's the point in defining an additional conformance class, if >> it doesn't make any difference in practice? (Methinks this is exactly >> what you said yourself just a few messages ago). > > To allow for extensbility of the language in the future. By discouraging > people from using non-conformant constructs we can in the future > introduce some additional parsing rules for new elements, for instance. I think if we really want to discourage people, we need to do more than flag that in conformance checkers. Remember: most people don't use them. > This is comparable with how 'color:foobar' works in CSS. foobar is > currently not a conforming value for the 'color' property. This > discourages people from relying on 'color:foobar'. In the future the > 'color' property could get a value 'foobar' that does something new in a > compatible way as older user agents will simply ignore it. I don't get that example: "color:foobar" (as far as I understand) just has no effect, so if people *meant* to set a color, they will fix their CSS. Contrary to that, there doesn't seem to be a reason for authors to produce conforming HTML5, if conforming processors are required to accept tag soup anyway. If the tag soup does what the author intended, why would she/he change it (unless she/he belongs to the small group of authors who already believe in producing conforming documents?). > The other advantage of keeping the conformant class smaller in this > particular case is to encourage authors to write documents that are > backwards compatible with older user agents. Because relying on some of > the specific quirks in the HTML5 parser algorithm might not guarentee > that. The error handling has evolved over the years and some browsers > may not recover in the same way as others currently. Playing the devil's advocate: in which case the best enocuragement for authors to produce conforming documents would be if UAs did *not* use the HTML5 parser, making it essential for authors to produce conforming docs in the first place. Best regards, Julian
Received on Friday, 4 May 2007 11:48:44 UTC