Re: 'Re: [xml] "Next XML schema implementation steps (poll)"'

[Kasimier Buchcik <kbuchcik 4commerce de>]

Hi,

on 8/16/2004 11:32 PM Daniel Veillard wrote:
> On Mon, Aug 16, 2004 at 10:52:41PM +0200, Torkel Lyng wrote:
>
> > What features are the XML-Schema part missing really?

Hmm, this sounds like we need a detailed description of the state of the 
implementation.
The question was related to missing points that were encountered by 
users and probably not communicated. Although I think that XML Schemata 
are not the ultimative way to go if the need for validation arises, one 
might tend to look for alternatives if a wanted feature - possibly easy 
to add - is not found, *without* ever notifying someone. See, I never 
sended a bug report to ms, and I don't know a single person who does. 
Especially the area of XML schemata seems to waft in an uncomfortable 
state, since the existing engines still differ in interpretation of the 
spec, thus creating a kind of distrust, related to the whole XML schema 
idea. You don't like to support something you don't like. I'm aware of 
some of the disadvantages of the spec, but we really need you to forget 
those accustomed prejudices (at least for the duration of the 
implementation ;-)).

>   Kasimier is the best person to answer, but I knw we are missing
> at least:
>   - content checking in derivations (I would have to implement 
>     comparison at the regexps level first)
>   - identity constraints not based on ID/IDREF
>  there is probably other parts though (do we have wilcards ?)

Yes, derivation of wildcards and validation with those is implemented 
('any' & 'anyAttribute').

>   To me one of the important things somewhat missing is more
> systematic regression testing for the code in CVS, I know Kasimier
> have been using at least subset of the W3C testsuite [1] but
> this is not integrated in CVS right now (though the full set of
> the W3C Schemas types are tested though Relax-NG James Clark
> testsuite). 

I agree that creating a safety buffer by integrating the W3C tests is 
more important than rushing for new features. On the other hand, and 
this would probably delight Robert Sanderson, this depends on the 
existence of the Python bindings. I'm would like to learn Python to 
create the tests if someone is willing to implement the bindings.

Greetings,

Kasimier