[tei-council] MD chapter revised: namespace rules

[James Cummings]

Sebastian Rahtz wrote:
> no, I am saying that Johann might prefer to see it this
> way, to avoid the chop and change of namespaces
> every other line. His choice.

Oh that's fine then.  So he produces a non-conformant document which he 
can then transform back at some stage should he desire.

> yes, whichever way you do it, conformant documents
> are involved. the distinction is between conformant
> and interchangeable, ie where the <altIdent>s have
> been reversed. 

Yes, I see what you mean.

> all conformant documents have
> the potential to be canonical as well, after a transformation.

If canonical only means that renamings are reversed, then this is true. 
  But remember there will still be all those extensions in the 
user-defined namespace(s) which can't be transformed back (reliably) to 
PureTEI (tm).


> So how about Roma shows you a new paradigm?
> Instead of "renaming div1 to section", it
> only  lets you add a new element called "section";
> however, when you create it, you have the short-form
> option of saying "I want it to be a *move* of div1,
> but I want to change the name,
> the description and the examples". So you get
> the "edit element" screen, but with no chance
> to change classes or content model; you are
> *forced* to use a new namespace.

This seems reasonable to me.  Would I also have a short-form of saying I 
want my element not to be a *move* but a *copy*?  I.e. given me the same 
content model/class affiliations as element X?

> This removes the possibility in webRoma (command
> line Roma is just an ODD processor, so not
> relevant) of doing simple renames. Internally,
> we manage the "move of div1to section" as
> a change of div1 and an <altIdent>.

Seems reasonable.

> I am groping here towards a work plan
> for rewriting Roma. 

Yes, I recognise that, and think there will be a lot more work here than 
perhaps originally conceived.

> If we can make the use
> of <altIdent> much harder, that strange
> case which bothers us arises less often; either
> way, a new namespace for a moved or new
> element is mandatory; and the algorithm for that
> is that we start off forcing you to think one up,
> and thereafter propose the same as you said
> last time.

Sounds good to me, I'm sure Johannes would be happy with that, don't 
know about Wohann though, but he's doing his own thing.

> Whether newly added attributes are forced to a
> namespace, as opposed to the default null one,
> is tricky. I am personally inclined to think
> that this is <soCalled>political
> correctness gone mad</soCalled>.

Current consensus seems to be that they should so they don't collide 
with existing elements, etc.

> I still believe that we have an option in Roma
> (maybe just on the desktop version for advanced
> use) which ignores the <altIdent>s
> and generates a script to reverse them and
> <equiv>s you have stuck in, probably by hand,
> in documents.

That would be nice. (But 'nice' taking less priority than other Roma 
development I think.)

-James