[sc34wg3] Feedback on TMCL draft 2009-10-19

[Lars Marius Garshol]

* Robert Cerny
> 
> 1) http://www.isotopicmaps.org/tmcl/2009-10-19/#sect-subjectidentifier-constraint
> The example would allow the empty string to be a valid subject identifier. Is the empty string a valid subject identifier?

The empty string is not a valid subject identifier according to the TMDM, which requires a full URI, so there is actually no way to get an empty string into the topic map in the first place. If that had been possible it would of course have passed validation.

> 2) http://www.isotopicmaps.org/tmcl/templates.ctm (dated  2009-10-19)
> I think the name "constrained-statement-type" would be more appropriate than "constrained-statement". In statements of this type, the players of the role "constrained" are statement types and not statements. Same goes for: constrained-role, other-constrained-role. "constrained-topic-type" sends the right message, IMHO.

You do have a point here, but the cost is of course that all the PSIs become longer. I created an issue for this:
  http://projects.topicmapslab.de/issues/1784

> 3) http://www.isotopicmaps.org/tmcl/schema.ctm (dated 2009-10-19)
> a) The schema allows types (e.g. topic types, name types) only to play the role "containee" one time at most. Why should a type not belong to more than one schema?

Good question, and one which we should perhaps consider more carefully. I created an issue for it: http://projects.topicmapslab.de/issues/1787

> b) The meta schema included in the draft is missing the statements about tmdm:subject. schema.ctm does not miss them.

You're quite right. Somehow these two are out of sync. I'll fix that when posting the next draft.

> c) The headline "UNIQUE VALUE CONSTRAINT" does not match the name of the constraint "unique-occurrence-constraint". If the role-combination statements are correct, the headline is right and "unique-value-constraint" would be correct.

Hmmmm. There is no unique-occurrence-constraint, so I'm not sure where you got this from.

> 4) Variants are left out completely. If that stays, then "occurrence-datatype-constraint" could be renamed to "datatype-constraint".

It could, but the current name seems more precise, and avoids problems if we were to add support for variants at some point.

> I discovered some errors (syntactical and semantical) in templates.ctm and schema.ctm. Additionally i must take this opportunity to speak for the rights of role types. They have the right to a name! :-)
> Attached you find an archive containing the original and the patched versions which i could load into tinyTim.

There were some useful corrections here, which I have incorporated. We really need to get a new draft out soon, because there have been many more changes in schema.ctm since the 2009-10-19 version.

Thank you very much for this feedback! It's very useful for us.

--Lars M.
http://www.garshol.priv.no/tmphoto/
http://www.garshol.priv.no/blog/