Customising and versioning

["Stephen Green" <stephen_green@bristol-city.gov.uk>]

I've been out of touch with UBL lately (admin problems)
but I've been having fun trying to work out the W3C Schema
derivation techniques again, trying to explore what can be
done with the UBL 1.0 schema files compared with UBL 2
schema files. This was prompted by Ken's push for using
W3C Schema after all with subsets but I've been 
concentrating on the version issue primarily (I thought,
Get that one solved and the rest should be easy).

Well I seem to have got the hang of it enough to produce
some derived minor versions of UBL 1.0, UBL 2 prd 1 and
UBL 2 prd 2. All three presented different 'problems' (or
'requirements') due to the changes in the NDR, especially
related to whether namespaces change with minor
versions. The result, as far as I could work things out,
was that UBL 1.0 allows derivation with substitution
groups (despite that being at odds a bit with UBL NDR 1.0)
as does UBL 2 prd 1, whereas UBL 2 prd 2 requires use
of redefine (which I did manage to get to work nicely).

My conclusion is that UBL 2 may be more difficult to
customise by W3C derivation than UBL 1.0 since 
it may be difficult to use these techniques if UBL minor
versions already use redefine for UBL 2. UBL 1.0 
allows use of substitution groups for minor versioning
(though that may be academic), with limitations due
to non-total global design, which would possibly allow
further use of substitution groups for customisation
of such minor versions.

If UBL 2 doesn't use derivation for minor versions, it 
allows derivation using substitution groups as long as 
the namespaces are changed in the derived schema files 
(hence the effect of not allowing such namespace 
changes for minor official versions limits techniques to 
redefine or redeclaration, I find).

Customising a minor version of UBL 2 would be
improved if the namespaces changed if there was still
a desire to use derivation for such minor versions. It
would then be possible to create substitution group
minor versions which could be further customised by
substitution groups mechanics.

Extensions do not seem to have any negative impact 
on the above: redefine for minor versions of UBL 2
and substititution groups for customised, changed 
namespace versions are still possible. It might therefore
be feasible for more extreme customisation to add
ID/IDREF anchor points where needed in the UBL
document schemata using derivation and then use
these in an extension :-)

I haven't yet managed to do any customisation of
datatypes using these methods in earnest. Maybe
I'll try that next.

All the best

Stephen Green