[sc34wg3] XTM 2.0 topicRef - proposal for an erratum

[Lars Marius Garshol]

* Lars Heuer
>
> The section mandates that the IRI reference contains always a fragment
> identifier. [...]

This is one of many things in XTM 2.0 which never received the  
discussion they should have because of the way the standard was rushed  
through. The whole thing was a very bad idea, but ... well ... that's  
history now. :-/

> That means the following references are illegal:
>
> * <topicRef href="http://psi.example.org/puck-die-stubenfliege"/>

Perhaps it should be, given that this is a reference to a subject  
identifier?

> * <topicRef href="puck-die-stubenfliege"/>

This is a reference to a file with that name in the same directory,  
and so very likely to be a mistake. In fact, one of the most common  
errors in hand-written XTM files is leaving out the anchor first,  
causing a stub topic to appear here.

> I don't see a good reason why every IRI reference must have a fragment
> identifier.

Basically that if you are going to point to a topic in another XTM (or  
CTM) file you're going to have to use a fragment identifier to say  
which topic you mean. The only way around it is to store a topic in a  
file on its own. Or to auto-generate item identifiers without fragments.

> The fragment identifier is only needed if someone wants to refer to  
> a topic within the same topic map: [...]

Nope. Say you have foo.xtm, containing the topic "bar". The reference  
to it will be foo.xtm#bar.

> The limitation to IRI references with fragment identifiers carries an
> unnecessary burden, especially if someone wants to generate topic maps
> automatically.

Maybe. I made an issue for this so we can track the discussion:
  http://projects.topicmapslab.de/issues/1459

Anyway, now that you've heard the rationale, what do you think?

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