Re: Type-assignment in one pass

[Eric van der Vlist <vdv@dyomedea.com>]

"K.Kawaguchi" wrote:
> 
> Probably now you see why it is called "pattern" in TREX...

Yes. Sorry to be reinventing the wheel...

> And just in case you don't know, you may also want to see XDuce, which
> is based on ML (thus a descendant of Prolog)

Ok, will do !

> It's exactly doing what you're thinking, I guess.
> 
> > I meant that we have ambiguity as long as the criteria to choose between
> > alternatives are not specified.
> >
> > There appears then to be 2 approaches: you can (like W3C XML Schema)
> > forbid any construct that would be ambiguous or allow constructs that
> > would be ambiguous and remove the ambiguity by specifying the algorithm
> > to chose between the different possibilities.
> 
> Right. My stupid question is, what is the merit of removing ambiguity by
> rule?

It's needed if you want to attach a datatype (as requested earlier in
this thread)...

A validation becomes then a transformation producing a post validation
infoset and you need to remove any ambiguity on the result.

Thanks for your comments !

Eric
 
> regards,
> ----------------------
> K.Kawaguchi
> E-Mail: k-kawa@bigfoot.com

-- 
------------------------------------------------------------------------
Eric van der Vlist       Dyomedea                    http://dyomedea.com
http://xmlfr.org         http://4xt.org              http://ducotede.com
------------------------------------------------------------------------