W3C home > Mailing lists > Public > public-html@w3.org > April 2007

Re: Version information

From: Ian Hickson <ian@hixie.ch>
Date: Thu, 12 Apr 2007 23:08:54 +0000 (UTC)
To: Eric Daspet <eric.daspet@survol.fr>
Cc: public-html@w3.org
Message-ID: <Pine.LNX.4.62.0704122305130.13484@dhalsim.dreamhost.com>

On Fri, 13 Apr 2007, Eric Daspet wrote:
> 
> SGML and XML serialisation are relevant here in my humble opinion.

SGML's isn't (nobody has ever used it except the validator).

XML's is relevant, but XML has its own HTML5 serialisation with its own 
rules, so isn't directly relevant to this discussion about text/html.


> If we use the "short" doctype <!DOCTYPE html> we will confuse people who 
> are used to xml, sgml and/or html4.

The group of authors that would be confused is far, far outnumbered by the 
group of authors who would find the shorter DOCTYPE a significant 
improvement and simplification.


> The doctype will not appears "valid" or "as usual" for them.

They'll get used to it.


> We will have a hard time to explain why we have two (or more) possible 
> doctype, one for the html serialisation and one for the xml/sgml 
> serialisation. Again, we will confuse people.

You can use "<!DOCTYPE html>" in both the XML and text/html serialisations 
and it works in both and is conformant in both.


> *IF* we decide we need a doctype, we need to have a doctype which will 
> be exactly the same in all serialisations. Therefore we will need a 
> doctype valid in sgml as in xml.

"<!DOCTYPE html>" is such a DOCTYPE. It is the proposed DOCTYPE.


> But, again, why do we really want to require a doctype in the document ?
> - Doctype may be usefull to SGML but we trash SGML as a need and no 
> modern browser use the DTD.
> - Doctype may be usefull to versionning but you advice not to version 
> html.
> - Doctype may be usefull to editors but a schema / relaxng will be 
> better and a direct support by the editor is certainly the way to go 
> (and not a dtd expliclitly in the document).
> 
> I left only one usage : versionning. Doctype switching is a kind of 
> versionning. Triggering a "new standard mode for html 5+" is also some 
> kind of versionning (and we will face the same problems for html 6, so a 
> "new new standard mode" will be needed in future).

Right, triggering "standards" mode in legacy browsers is the only reason 
HTML5 has a DOCTYPE at all.


> Is doctype the better solution for versionning ? A simple "version" 
> attribute in the <html> tag is simplier and does not interfer with 
> serialisations. We can even name it "standard-mode" instead of "version" 
> if we really want a single shot switching and not a versionning.

We're somewhat limited by legacy browsers when deciding what DOCTYPE to 
use.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'
Received on Thursday, 12 April 2007 23:09:03 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 9 May 2012 00:15:53 GMT