[Fwd: Re: [tei-council] on conformance document

[James Cummings]

Sebastian Rahtz wrote:
> James Cummings wrote:
>>
>> At what point in changing the TEI with an ODD does my schema not
>> become kosher?  i.e. What can I do with ODD that makes my schema
>> non-kosher?
>>
>> The way I understand your reading is that:
>>
>> 1) I rename some elements = kosher
>> 2) I remove some elements = kosher
>> 3) I add some elements = kosher
>> 4) I change some classes of existing elements = kosher?
>> 5) I add elements from other namespaces (e.g. SVG?) = kosher?
>> 6) I rename tei/teiHeader/text elements = not kosher?
>> 7) I don't have (however named) title/publication/source description =
>> not kosher?
>> 8) I create the root element in a different namespace = not kosher
> which confusion was precisely why I tried to introduce degrees of
> kosherness,
> and explain the rules. In fact, "kosher" is NOT the right parallel, as
> that's
> a yes/no distinction. Syd is (possibly) arguing that any schema for which
> there is a valid TEI ODD is TEI-conformant. Pragmatically, I simply
> cannot accept that. TEI documents which conform to a schema
> meeting only 1) and 2) above are incontrovertibly "better" TEI
> than some others, in my opinion; mainly because they are
> more interchangeable. Is the document
> <TEI><teiHeader>...</teiHeader><foo>hello world</foo></TEI>
> _really_ muchuse to man or beast in a world of other TEI documents?

In my view if something is a direct subset of tei_all and as such needs no
transformation to have it interoperate smoothly with other TEI documents, then
that is a good thing and a much better thing than a TEI document which needs any
form of transformation to make it interoperable with other TEI documents.  If
some syntactic sugar renaming has taken place then that is an easy
transformation that people with no specialised (e.g. XSLT) knowledge could do by
search-and-replace in an editor, but is still less convenient than something
which is a pure subset of tei_all.  If the TEI ODD creates new elements or
attributes which have no equivalences in the tei_all schema, and so can't be
converted back to be a tei_all subset without some creative tag abuse, then
although it is a TEI conformant document (in having an ODD) it is necessarily a
very different thing than a document which is a pure subset of tei_all.

This doesn't mean I'm in any way against changes which result in schemas that
are not pure subsets of tei_all.  Indeed, if we accept (and I think we do) that
it is more desirable for projects to encode things in a manner which make
semantic sense to the type of document they are working with than abuse existing
elements, then at least deviating from a standard base -- and more importantly
documenting that deviation with a TEI ODD -- is the best we can hope for in a
world necessary fragmentation.

However, this doesn't change the principle that documents which validate against
pure subsets of tei_all are 'better' for interchange and interoperability than
documents which don't.  Although I tend to view simple renamings (documented
with <equiv>) as basically the same as something validating against a subset of
tei_all, and have never been a fan of binary oppositions, I think it is true to
say that there are two types of schemas.  There are those which are subsets of
tei_all (where element choices or values have been constrained, etc.) and those
which are not.  Although both of these (if they have an ODD) are conformant
documents, for the purposes of interchange and interoperability they are
inherently different kinds of things.

I'll stop babbling, must have been all that time off in Tuscany,

-James