Re: [ubl] Discussion of substitution groups

["Stephen Green" <stephen_green@seventhproject.co.uk>]

My general belief here is:
Don't repent in the dark over what you
decided in the light.

That is, we had more participants and
a greater collective knowledge base when
these issues were all first decided. I don't
think the fact that the mechanism XSD has
on offer for implementing UBL's agreed
aim of inheritance happens to be
substitution groups changes the agreed
approach. Nothing has changed here.

I've just been reading the discussion
minutes for the "W3C Workshop on XML
Schema 1.0 User Experiences and
Interoperability"
http://www.w3.org/2005/03/xml-schema-user-program.html
and nothing suggests that it is a good idea to
keep the same namespace for minor versions,
only the opposite.

Apologies I can't give details on my previous
involvement which showed there are problems
with keeping the same namespace for minor
versions. All I can say is you end up wanting
to find any way you can to show in an instance
what minor version is used and the namespace
just confuses things by implying that two are
from the same schema even though they aren't.
You end up with a version element in the
content!

Comparison with CEFACT at this stage seems
likely to mislead in that the ATG2 schemas
are 'local' and therefore cannot use this
extension mechanism (though I'd argue that
that all the more should lead to a change of
namespace for any version change and suggest
that 'minor' is meaningless since it depends
on a concept of backwards compatibility which
cannot be substantiated normatively.

In all I think it would be a bad idea to go back
on what was decided by a pool of experts in
UBL with greater collective knowledge and
is backed up by observations such as those
brought to the June W3C workshop noted
above. (Noting the desire expressed to allow
derivation use for local schemas - even those
using 'local' wish to use derivation, probably
with substitution groups, but can't.)

All the best

Steve


----- Original Message ----- 
From: "CRAWFORD, Mark" <MCRAWFORD@lmi.org>
To: "Stephen Green" <stephen_green@seventhproject.co.uk>;
<ubl@lists.oasis-open.org>
Sent: Tuesday, July 19, 2005 11:48 AM
Subject: RE: [ubl] Discussion of substitution groups



 > I remember we discussed it in Hong Kong
> and decided we need the minor version
> namespace changes.

But when we discussed this in Hong Kong, we didn't talk about
substitution groups being required to support it.

> I have seen the problems you get when
> the minor versions don't have namespace
> changes and they are highly regretable. I'd
> hate to think of not learning at all from
> history.

Please provide details.

> Then of course there is the option that we
> don't have minor versions and that seems
> a catastrophic departure from all UBL has
> achieved so far (IMO).


I am not suggesting we don't have minor versions, only that namespace
versioning of the minor versions is an unnecessary overhead that we
should avoid.  We have eliminated this in CEFACT for this and other
reasons.  Also remember, we are a model driven solution.  We have a
library that details our content models.  When a BIE changes, a new BIE
is created.  Why should we have different type definition rules for the
derived BIE?  Exactly what is gained by using type derivation rather
than just defining a new type that replaces the old in our schema?

Mark