[tei-council] solving the Birnbaum Biznai

[Syd Bauman]

> what aspect of <msIdentifier> is not handled?

I phrased that poorly. Of course <msIdentifier> *can* be handled, but
not as a single class, no matter how clever we make classes, they're
not content models.

The content model of <msIdentifier> has structure:

  element msIdentifier {
     att.global.attributes,
     (
        (
           (
              country?,
              region?,
              settlement,
              institution?,
              repository,
              collection?,
              idno?
           )
         | msName
        ),
        ( altIdentifier | msName )*
     )
     }

Note that in the inside set there <settlement> and <repository> are
not optional, but everything else is.


> If you add a dependency in "msdesc" on "namesdates", but the user
> then deletes <settlement> from namesdates, you are no better off
> than you are now. Module dependency is a crude tool. That's the
> main reason I don't want to use it.

Ah, now I understand. This doesn't bother me because it is no
different than what happens when the user deletes a required child
element when the required child element is in the same module as the
parent. (It doesn't mean I don't think we should try to do something
nicer, it just means it's orthogonal to module dependency.)


> sure. if you define what the precise behaviour of module dependency
> is.

I'm not sure how you're defining "behaviour" here. Is something
like "if a module depended on by an explicitly loaded module cannot
be found and loaded, an error condition should be raised and no
output (schema, reference doc, etc.) should be generated"
insufficient?


> that's module dependency...

Well, yes, it's module dependency merely documented instead of properly
tool-enforced. 


> I don't disagree with that choice. But it leaves the TEI making
> invalid schemas rather easily, as James found.

Right, which is why I'm not so fond of it, and far prefer your
suggested 3.5 solution.


> surely, I can do that. its just "people who bought this module also
> bought module".

I'm imagining a little more forceful than that. I'm imagining "when
you buy this you get that for free" ... if you don't want 'that', you
have to explicitly say so.


> I think I'd agree; but would note that supporting it there is
> probably not much more work than in schema.

If you say so. Sometimes makes my head spin thinking of the stuff you
do. Although it's far worse today, as my head was already spinning
from jet-lag. :-)