NDR 2 Comment / RE: Customising and versioning

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

Background:

I feel strongly that the business users of UBL would benefit
by making the subsetting of UBL without a change of the
namespace the prefered customisation technique, where
possible. For this there are the following techniques as far
as I'm aware which seem to be usable with UBL 2:

To describe a subset without the need for a namespace change:
1. CAM files
2. XPath files
3. W3C XML Schema redefine
4. W3C XML Schema rewriting of entire set of schema files

All these can contain IDs which can be referenced from an
ebBP definition as part of the description of the document(s)
used for the process (ebBP 2.0.3).

Each technique has a different set of benefits related to different
use case requirements and so each has strengths and weaknesses
depending on the target scenario.

Related to UBL 2 NDR

There might be an impact of the yet-to-be-chosen mechanism
for UBL TC creation of minor versions of UBL 2 on the above
methods of subsetting and other customisations (such as
extension). The use of redefine to create minor versions might
limit the number of further uses of redefine (W3C XML Schema
'xsd:redefine') due to the extra stress palced on XML tools 
such as parsers and validators on following the complexities of
multiple redefinitions. Personally I found that the following works
with the tools I use:
1. use redefine to create a minor version by derivation from a
previous major version of UBL 2 schema files
2. use redefine again to designate the extent of a subset
3. use substitution groups to extend the subset schema set
but there were the beginnings of possible cracks appearing
in the validation once the extension was added. Although this
was anecdotal evidence in that there could have been any
number of reasons for the slight flakiness in validation at the
end, the complexity might be a factor in limiting the actual
extensibility of such subsetted minor versions if redefine is
used for the minor versioning.

I don't see this as a sufficient reason to not use redefine for
minor versioning but I would like to see the decision made about
this and this added, perhaps in an appendix to the UBL 2 NDR
or to the main NDR standard before it is standardised because
it will be a factor in planning implementations of UBL with regard
to techniques to be used in subsetting and other customisations.

Many thanks

Stephen Green