I'm sure there were due considerations given to the idea of
two schemas (A & A' using your terminology) per UBL-schema.
I'm just trying to put the rule into perspective now that
we have some form of draft schemas to talk about.
Say, A' is the one with documentation. People would and
could strip documentation away from A' to get A, where
A = A' (modulo documentation elements that
do not affect processing in any way)
So the part about "*and* change the namespace" does not have
to necessarily take place, unless there is a simultaneous
contextualization into the user's operating environment.
There's no tons-and-tons of documentation now for the draft.
But that possibility could exist in user environment, especially
for contextualization purposes, they wish to document differences
made, etc. So when that ton-and-tons of documentation do
surface, perhaps user could "point" from their schemas using
URLs to a full-fledged page to document that point in their
schema.
Complete documentation cannot be indefinitely stuffed
into schema as if schema is the best way to store documentation.
So the "tons-and-tons of documentation" scenario ought not
to happen (even though it can), and may be even stated as
an NDR checklist item as a form of best practice.
Stripping documentation as part of optimization does not warrant
a change of namespace since the stripping may be carried out in
various programming styles, such on-the-fly, cached, pre-compiled
binaries, etc. The way UBL annotates the element using
xsd:annotation, it also falls outside of schema checkers,
so that there is no need to modify schemas and change namespaces.
As a result, there is no need to specify the maintenance of
two identical schemas (A & A') where they differ only in
annotations.
Best Regards,
Chin Chee-Kai
SoftML
Tel: +65-6820-2979
Fax: +65-6743-7875
Email: cheekai@SoftML.Net
http://SoftML.Net/
On Mon, 30 Jun 2003, Eduardo Gutentag wrote:
>>Although I agree with you in principle, the problem is that if
>>we were to produce enormous schemas with tons and tons of
>>documentation embedded, carrying some ubl namespace, people who
>>wanted to use them with any kind of hope of acceptable
>>performance would have to strip the documentation away *and*
>>change the namespace.
>>
>>It was because of this concern (that is, allowing people to say
>>they use UBL schemas) that we came up with the idea of two
>>sets of schemas carrying the same namespaces names (that is,
>>schemas A and A' being both in namespace UBL:A if the only
>>difference is documentation.) Perhaps we should play with the
>>idea of having A in namespace UBL:A and A' in namespace UBL:A' ?
>>
>>(where UBL:A is just a shorthand reference to the UBL namespaces,
>>not to be taken as implying that it is an UBL namespace name...)
>>
>>Chin Chee-Kai wrote:
>>> Again, this two schemas per model rule is also something
>>> I feel is rather stringent to be stated as a rule, or that
>>> it is redundant.
>>>
>>> Developers and users will find their own most suitable form
>>> of optimizing for processing. It shouldn't be a specified
>>> form of rule for such purposes.
>>>
>>>
>>>
>>>
>>> Best Regards,
>>> Chin Chee-Kai
>>> SoftML
>>> Tel: +65-6820-2979
>>> Fax: +65-6743-7875
>>> Email: cheekai@SoftML.Net
>>> http://SoftML.Net/