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

On 15. apr. 2007, at 15.00, Henri Sivonen wrote:

>> Is it possible to specify all  machine-checkable criteria in  
>> XHTML5  by means of:
>> A RELAX NG schema which refers to
>> 1. A datatype library
>> 2. A set of formal rules, specifying, for instance, table  
>> integrity requirements
> A RELAX NG schema cannot refer to a generic "set of formal rules".  
> I take it that you meant to ask if XHTML5 conformance can be  
> defined as an XML document satisfying a RELAX NG schema and a "set  
> of formal rules" (without the schema actually doing the referring).
> In theory, yes, but the "set of formal rules" would just be a catch- 
> all notion for everything that the RELAX NG schema doesn't cover.  
> To implement a conformance checker, you eventually need to have  
> running code that checks for all the stuff that the RELAX NG schema  
> didn't cover. With the current WHATWG spec model, a human (me)  
> takes English as the input and produces a computer program as the  
> output. You seem to be suggesting adding an intermediate formalism.

What I'm trying to find out here, is wether or not it is possible to  
produce a specification that is based on more formal frameworks than  
english prose. We don't need a single all embracing formalism to do  
that. So fare It looks like we need something like the components  
I've outlined above.

Among other things, this will make it much easier to write and verify  
applications like your conformance checker.

I also think that a lot of competent and standards aware developers  
will appreciate a specification that draws on well established  
notations and languages .

.. And I think it is fair to think it might perhaps also induce some  
more rigor into the standard  :-)


Received on Sunday, 15 April 2007 14:51:53 UTC