[dsdl-discuss] Re: Is SGML in scope for DSDL?

In response to my:

> b) Restricting ourself to XML 1.n features is not feasible as we have
> already identitifed features DSDL needs that XML 1.n will not define. In
> addition, we will need to be able to conform to XML 2.n, 3.n, etc, so
> writing a restriction to conform to XML 1.n into the spec will be a
serious
> inhibition to the long term use of DSDL

Rick asked
>Which are these?

Among others:

The ability to make changes to models that are dependent on the contents of
elements. (Remember XML 1.n includes the use of DTDs as well as the rules
for well-formed documents!)
The techniques we develop to identify choreographies associated with
instances (unless we force them into PIs.)
Element renaming

I also wrote

> c) Restricting ourselves to the XML Infoset is also irrational when we are
> going to provide information that is not currently included in the Infoset
> (e.g. characters subset info). Certainly we should ensure that all of the
> Infoset if covered, but to restrict ourselves to this minimalist approach
is
> just not acceptable

To which Rick replied
>I don't understand this. The character subset is a constraint against which
we test documents (and or schemas) surely?

Yes, but surely we need to be able to pass on information about which
character set the element/attribute has been validated against as part of
the enhanced infoset created by the validation process. To suggest that we
should validate and not enhance the information set as a result seems
illogical to me.

When I wrote

> d) We have agreed that we need to be able to make the extensions such as
> datatypes and cardinality available to those using DTDs in preference to
> schemas (the vast majority of current users). If we extend DTDs then the
> extensions should be designed to work with SGML DTDs as well as with the
> subset of SGML DTDs adopted for XML 1.0

Rick resoponded
>I thought we had no editor for that. I think I sent around a note showing
that
there could be a binding of XML Schemas using SGML DTDs, that could be
the basis of such a thing. But I really think this is an 80/20 issue for
SGML:
it would be enough merely to show how datatyping can be done using
the DATA keyword on declarations, and show how to reference other specs
in a RELAX NG-like fashion.

We'll find an editor for this when we are ready to do it. We cannot,
however, just "show how datatyping can be done": we need formal productions
to extend both 8879 and the XML DTD spec.

Martin

--
DSDL members discussion list
To unsubscribe, please send a message with the
command  "unsubscribe" to dsdl-discuss-request@dsdl.org
(mailto:dsdl-discuss-request@dsdl.org?Subject=unsubscribe)