Re: Formal definition of HTML5 (was Re: Version information)

On Fri, 13 Apr 2007, Henrik Dvergsdal wrote:
> On 13. apr. 2007, at 11.05, Ian Hickson wrote:
> 
> > > How will the HTML5 language be formally defined?
> > 
> > It's currently formally defined using English prose.
> 
> But there will eventually be a formal language definition right? I mean 

Anyone is free to implement formal language definitions, but I see no 
reason to make one more "official" than any other. Reference 
implementations are often a source of bugs and constrain the development 
of the specification in ways that are artificial and unrelated to the 
needs of the users and authors ("we can't require that, the grammar 
couldn't express it").


> - you cannot validate documents against english prose.

Sure, you just have a conformance checker that implements the prose. For 
example:

   http://hsivonen.iki.fi/validator/html5/

There's no need for an officially blessed non-prose formal language 
definition to do that; indeed having one would likely introduce problems 
-- for example, DTDs have caused a long legacy of people "validating" 
their HTML4 documents while missing entire classes of conformance errors, 
because DTDs are completely unable to ascertain conformance of many things 
(e.g. complex attribute value syntaxes).

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Friday, 13 April 2007 10:23:12 UTC