Re: Defining data in reverse chronological order

Charles Baines asks:

> Which returns to the question - can <sequence> define
> the order of elements by an attribute value common to
> each element without having to physically list the
> elements in the order required?

No, <sequence> constrains the order based striclty on
element name.  There have been proposals to add various
forms of what are collectively called co-occurrence
constraints, though I haven't seen many requests for
sorted sequences of the sort you suggest.  Of course,
there is no end to what users might want...prime numbered
entries first?  Sounds bizarre, but a mathematician
might want it.  Unless you're willing to move away from
a declarative language to something Turing complete
(I.e. a full programming language for the constraints),
you'll always wind up trying to make reasonable tradeoffs
between complexity in the schema language and its 
ability to express complex constraints.  I think many
of us have a very strong preference for retaining a 
language that's basically declarative, as it is so
much easier to reason about and build tools for
such languages.

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------

Received on Friday, 13 June 2003 16:45:19 UTC