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

[Sebastian Rahtz]

James Cummings wrote:
> Because your subject line relates this to conformance, I'm assuming 
> that your implication is that we should reconsider the requirement of 
> ODD for conformance?  Does the fact that this is fairly easy, and has 
> been done previously, affect the council's decision that an ODD is 
> required for conformance?
not really. I am generally happy with TEI Conformance being a serious 
target,
including ODD (though see below).

I do want to remember the debate, though, about what we should deliver.

It seriously affects the description of what an ODD processor should do if
I say that it must generate a parameterized schema with clearly defined
entry points. Can the user guarentee that for every "model.foo" she reads
about in the guidelines there is a corresponding RELAX NG pattern 
"model.foo"
in a RELAXNG schema published by the TEI? there is no apriori reason
why R_ODD (ie the XSL stylesheets which are beyond Roma, and are our
defacto definition of what an ODD processor should do)
_should_ preserve those names, you understand, or even preserve the 
knowledge
of classes in the schemas it generates.

In the flurry of emails over Easter which I finally read (don't you guys 
have
holidays?), I see a pretty clear consensus that conformance and ODD are 
closely
linked, and that modification via DTD or RELAX NG fragments is simply not
on. I am not sure I see a clear conclusion as to whether we continue to 
generate
the DTD and RELAX NG modules for those who don't give a dang about
conformance.

Can I introduce you to someone?  This person (let's call him Wohann :-})
shuns ODD entirely, and maintains his customization by writing
a wrapper RELAX NG schema which does just what he wants, pulling
in bits of TEI RELAX NG schemas. It is readable,
documented, efficient and expressive. He uses extra features of RELAX NG
or other ISO DSDL standards to manage his documents to a high standard.
He is, you may say, non-conformant. He cannot show his "workings"
in the approved form, and gets an F.

HOWEVER, when you get Wohann's document,
it validates fine against tei_all. If he could be bothered, he could write
an ODD equivalent of his hand-crafted schema, and the document would
validate against that too. The document validates against _many_ schemas,
and breaks no TEI semantic or syntactic rules. It adds no new elements
or does any renaming, or breaks any models.

So two questions:

 a) Can Wohann put the "TEI Conformant" badge on his web site, by the 
fiction
that his ODD is the implicit tei_all? Even though he and all his workers
use the evil little hand-crafted schema, and the document contains a PI
linking to that?

 b) do we encourage/support/allow Wohann's working method by generating
the schemas for him?

Or I am just going the mulberry bush again?


-- 
Sebastian Rahtz      

Information Manager, Oxford University Computing Services
13 Banbury Road, Oxford OX2 6NN. Phone +44 1865 283431