[tei-council] Conformance .... the continuing saga

[James Cummings]

Syd Bauman wrote:
>> Should we actively support non-Conformant methods of doing things,
>> or simply have a short notice that there are other ways to do it
>> with some descriptions?
> 
> I think this is analogous to use of start= of <schemaSpec>. If you
> specify <schemaSpec start="div">, you end up with a schema that will
> only mark as valid documents that are definitionally non-conformant
> (because they have no header). The system supports doing so, because
> it is often a *really* big help when making test files, etc. But it
> isn't part of a TEI Conformant system. Similarly, you can stitch
> together a schema from Relax NG fragments when it is more convenient
> to do so (you're experimenting, you're on an airplane w/o access to
> Roma, etc.). But the result is non-conformant, and we shouldn't spend
> a lot of effort (if any) describing how to do this.

Exactly my thinking.  We should maybe describe that it can be done and
might be useful in those circumstances, but I wouldn't suggest more than that.

>>> (a) we are currently generating and distributing them
>> Not a good enough reason to keep doing so
> 
> Agreed. IMHO distributing parametrized DTD fragments is not only a
> waste of Sebastian's time (although he points out not much), it is
> potentially confusing to a DTD user who hasn't read the Guidelines
> carefully, but remembers extending P4!

Exactly.

>>> (b) we haven't told anyone we intend to stop doing so
>> Political, but again, P5 is such a big change
> No problem. We'll tell 'em now. (Or at release 0.7.)

Yup, along with the not-adding-elements to the TEI namespace thing. ;-)

>>> (c) people e.g. wendell might want to use them
>> Those who don't care about Conformance? And nothing to stop them
>> using that locally, but producing versions that do the same thing
>> in an ODD.
> Wendell will want to stitch together Relax NG, sure, but who wants to
> use parametrized DTDs? Remember, doing so is going to give you all
> sorts of namespace headaches.

I certainly won't -- what are some use cases for doing so assuming one can
use ODD to do the same thing?

> Whether we add Relax NG info or not, we should remove stuff about
> using parametrized DTDs. It's a bad idea.

If we aren't supporting it then I say remove it (except maybe a note about
the change and how it used to be done).
> Perhaps a discussion of how to stitch together TEI Relax NG schema
> fragments should be the subject of a white paper, and not in the
> Guidelines? 

Isn't this what the now-conference-like bit of the TEIMM is for?

-James


-- 
Dr James Cummings, Oxford Text Archive, University of Oxford
James dot Cummings at oucs dot ox dot ac dot uk