- From: Robert J Burns <rob@robburns.com>
- Date: Tue, 22 Jul 2008 18:58:00 +0300
- To: Smylers <Smylers@stripey.com>
- Cc: public-html WG <public-html@w3.org>
Hi Smylers, On Jul 22, 2008, at 4:17 PM, Smylers wrote: > > Robert J Burns writes: > >> ... section 2.1 says that a Doctype is not required by the draft >> while >> section 8.1 says that it is required by the draft: both >> normative. ... >> a text/html serialized HTML5 document either does or doesn't >> require a >> doctype depending on whether the author follows section 2.1 or >> section >> 8.1. > > The only mention of "doctype" I can see in section 2.1 is: > > Such XML documents may contain a DOCTYPE if desired, but this is not > required to conform to this specification. > > Clearly that's talking about XML, so isn't relevant to text/html. I'm not sure that's clear at all. Why would we be discussing the optional use of a doctype other than HTML5's doctype? That's not our business. Obviously anyone using XML is permitted to use any doctype they want. They do not need the HTML5 recommendation to tell them that. But why not just make this simple and say the doctype for HTML5 is "<!DOCTYPE html>" (case insensitive for text/html and case sensitive for XML) and be done with it? Why are we introducing all of this confusion into the recommendation? What problem statements do these complications address? In any event, if we require the doctype for the text/html serialization and the XML recommendation requires it for our HTML5 XML serialization then why even mention it as not required to conform to the HTML5 specification. It simply adds in some obscure esoteric language and introduces an unnecessary difference between the two serializations. Others off-list argued that section 8.1 requiring the doctype only applies to the text/html serialization, but it doesn't make that explicit. Section 8.2 specifically starts: "his section only applies to user agents, data mining tools, and conformance checkers." If this is true of all of chapter 8 why wouldn't that appear at the top of chapter 8, or repeated for each section for which it applies. It also would be clearer if we didn't try to use the term HTML (a fairly central term in our endeavor ) to mean both HTML (our specification) and the text/html serialization (one of the serializations of our specification). To recap, if HTML5 requires a doctype for the html serialization. And XML requires one for our XML serialization (and the same one would do just fine). Then why introduce this difference. What use case does that difference address? Are we hoping that the next XML rev drops the doctype requirement? Again, this has taken us away from the original discussion of named character references which do not have anything to do with this doctype discussion. Except they have something to do with it because we're inviting authors to use a doctype that is not specific to HTML5 (or HTML at all) declare their HTML document type: which strikes me as very wrong. For an XML implementation to process HTML5 it has to know when it is encountering HTML5 (or any similar HTML serialized as XML). Without a mechanism to do that, the processor will not know it is HTML. In documents conforming to the Namespaces in XML recommendation, the HTML namespace URI provides that information to the processor. In non- namespaced documents or in non-namespace aware processing applications, there has to be another mechanism. One mechanism might be that such processors will always encouter the doctype "<!DOCTYPE html>. The XML schema definition recommendation provides an other means, but it is somewhat dependent on namespaces already. That leaves only the doctype to indicate that the processor is processing HTML. So how can we leave that as something for other recommendations to worry about? How does an XML processing application know that it is processing an HTML (specifically an HTML5) document (for the authors who leave of the doctype because we allowed them to do so)? Take care, Rob
Received on Tuesday, 22 July 2008 15:58:57 UTC