Re: deterministic content model

[Murali Mani <mani@CS.UCLA.EDU>]

I think I have to use some of our own terminology, probably not very well
known to lot of people, though I think it can be guessed. First
terminology is a regular tree grammar = an XML Schema.

I thought about this a little bit after we discussed about ambiguities in
regular tree grammars.
This is very tentative -- I think key constraints are specified in schema
by saying something like -- for all values of Book type, the key is so and
so -- have to verify this with XML Schema.
But when we have ambiguities, we do not know exactly what are Book types.
I think we should get around this a little bit -- there could be multiple
solutions -- first check whether we can ensure that when we have to
specify key constraints (I think results of document processing do not
need to specify key constraints), therefore I tentatively believe that we
need key constraints only for the initial data modeling part, then there
can possibly be multiple solutions -- either ensure that the DB designer
can specify only unambiguous grammars (is this solution good enough?) or
warn the designer, or set a rule for handling ambiguities. Other solutions
also could be possible.

This needs some more concrete work, and I am very thankful for your input
on ambiguities in regular tree grammars -- I think it is *very* useful.

I think what we need for handling key constraints easily is non-ambiguous
regular tree grammars, but XML Schema is a *very* small subset of
non-ambiguous regular tree grammars.

<warning>speaking for himself only<warning>

regards - murali.

On Tue, 17 Apr 2001, Kohsuke KAWAGUCHI wrote:

>
> > Tentatively, deterministic content models are easier for specifying key
> > constraints, but I think document processing (such as XSLT or Query) gets
>
> I really appreciate if you would teach me how key constraint enforcement
> gets easier in deterministic content model.
>
>
> ----
> Kohsuke KAWAGUCHI
> Phone: 650-786-0721
>
>