- From: Robin Berjon <robin.berjon@expway.fr>
- Date: Mon, 04 Jul 2005 01:31:44 +0200
Ian Hickson wrote: > No, I expect typical authors to be quite happy not worrying about generic > XML authoring tools and just to use [ ] instead, which is perfectly > conformant in HTML5. I can't help but take exception with this kind of argument, especially coming from people who should clearly know better. For crying out loud the Web is *BIG*. Authors who fall into the "atypical" box could count in the hundreds of thousands, if not millions. Would a browser that just handles the typical Web document (whatever that is) be good enough? That would probably leave a good 30% of the HTML out there completely broken. Same goes for Web technology, while it can't cater to anyone it should clearly strive to take more than just Joe into account, no matter how nice Joe is. > Actually I expect most authors to never even remotely come close to this > problem, since you only hit it if you are doing _all_ of the following: > > 1. Using repetition blocks with the template name substitution feature. > 2. Nesting such templates. > 3. Using the XML serialisation of HTML5. > 4. Trying to use tools that are based on XML DTDs or XML Schemas. You don't need 4 to be true, and not necessarily 3 either. Whenever I write small and silly Perl scripts to handle HTML documents on a box the way Joe Sysadmin would I tend to use whatever's best in that toolbox, and as it happens a lot of those tools are just XML parsers that happen to also know how to parse HTML, even though they might not be trying to be completely compliant (but they're "on my box" tools, so why should I care? I don't want them to be as complicated as browsers are). It's likely that arbitrary incompatibilities will make at least some of them unhappy, for reasons unobvious to their users. IMHO no matter how many syntaxes HTML 5 has, they should all be able to encode the same information. Otherwise people will be having to chose between them for the wrong reasons. >>Personally, I prefer Matthew's idea to use a templateid attribute. > > As noted, that doesn't solve the perceived problem. If it's an attribute that's not of type ID it can contain anything it pleases, and if it generates an ID attribute at runtime, all that's needed is for the latter to be a well-formed ID. -- Robin Berjon Senior Research Scientist Expway, http://expway.com/
Received on Sunday, 3 July 2005 16:31:44 UTC