[tmcl-wg] Feedback on TMCL presentation in Montreal

[Lars Marius Garshol]

Dmitry presented the new TMCL at the ISO meeting in Montreal
yesterday. The following is my feedback on those slides. It was mostly
presented in the meeting, but only orally, and so this is just a
written summary to make sure nothing is lost.


The infoset formalism is used differently in TMCL from in TMDM. I can
see why, but I think one of them needs to change so that they are
consistent. (The definition of the properties is different, as is the
naming convention for properties (space in names vs dromedaryCase,
singular/plural), etc.)


The [schemaName] property on topic map schema: where is the rest of
the topic map? This is just a fragment, so what does the rest look
like, and how does it hang together?


I think the [include] property should go away. It's useful in
implementations, but has no place in the standard.


The [sourceLocator] property should be [itemIdentifier].


The OrTopicExpression item should be replaced by TopicSet.

Similarly, the SubjectIdentifierSchema and SubjectLocatorSchema should
be unified into just LocatorSchema.


Should the datatype really be a selector on variants? It's
inconsistent with how occurrences are handled. It also seems more
natural for it to be a constraint.


[oneOf] on occurrences just encourages bad modelling; we should take
it out. Someone who really needs it can use [assert], anyway.

Similarly, the [card*] properties on roles are somewhat dubious in the
sense that anything other than 1-1 and 2-2 is likely to be poor
modelling.


On role schema, what is the interpretation of [otherRoleSchema] when
it has multiple values?


Why is only AND merge of schemas allowed, and not OR merge? Also,
should TMCL define merging of schemas at all? If so, how and why?

-- 
Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no >