[sc34wg3] XTM 2.0 topicRef - proposal for an erratum

Lars Marius Garshol larsga at garshol.priv.no
Tue Oct 20 07:54:34 EDT 2009

* Lars Heuer
> I found more issues regarding XTM 2.0. XTM 2.0 enforces that topics
> have at least one item identifier (the id attribute) which is no
> requirement of the TMDM.

That's true. In fact, there was an XTM 1.1 version for a while which  
required topic elements to *either* have an -id- attribute *or*  
specify some form of identity inside the element. That disappeared  
when we made XTM 2.0, and frankly I can no longer remember why.

This is yet another of those issues which should have seen proper  
discussion, but never did.

I created an issue for it: http://projects.topicmapslab.de/issues/1460

Personally, I think requiring the -id- is OK, if not ideal, but other  
opinions would be welcome.

> Further, I think it was a mistake to remove the possibility to
> reference a topic by its subject identifier / subject locator. In XTM
> 1.0 we have the possibility to reference a topic by its subject
> identifier / subject locator, while XTM 2.0 enforces (again) item
> identifiers.

This was definitely a mistake. I've spoken to a big Ontopia customer  
which essentially had to give up on using XTM 2.0 because of this. And  
it's clear that in XTM fragments this is really awkward.

Graham and I agonized over this one for days, in the end coming down  
on the side of simplicity, partly for emotional reasons, I suspect.  
Again, without proper discussion of the draft, the problems were never  
spotted by anyone else, and this is what we ended up with.

I made an issue for this, too: http://projects.topicmapslab.de/issues/1462

Further input from anyone certainly welcome.

> I've attached a RELAX-NG scheme for a XTM version that is (differently
> to the proposal [1]) backwards compatible to XTM 2.0. Any XTM 2.0
> topic map can be parsed the the proposed XTM 2.1 version. [...]

Yes, this is pretty much how I'd solve the problems, too, if we were  
to make the changes you suggest and create an XTM 2.1.

The question is: is it politically wise to make an XTM 2.1? We already  
have two XTM versions, which IMHO is one too many. Should we really  
increase that to three? And what are the odds that there will be a  
proper discussion of the issues this time around?

Thoughts welcome.

--Lars M.

More information about the sc34wg3 mailing list