[sc34wg3] Question on TNC / Montreal minutes

Lars Marius Garshol sc34wg3@isotopicmaps.org
05 Sep 2002 16:50:38 +0200

* Lars Marius Garshol
| To be really precise: what we are talking about is the TNC. What you
| are talking about is what impact that has on the Reference
| Model. The representation of the SAM in the Reference Model needs to
| distinguish between base names that are labels and base names that
| are identifiers.

* Jan Algermissen
| Sure, but how does all this affect the reference model ? All you
| need to do is to provide the right PSIs (together with the
| explanation of their semantics) at the SAM level and then define how
| an XTM instance is to be represented as a topic map graph using the
| PSIs and the reference model.

Ah, you are saying that the TNC does not affect the RM, but just the
modelling of the SAM in the RM? If so, I agree.

| [...] 
| Hmm, if you agree on this, why is the TNC issue an exception ?

I don't think it is. We've probably just been miscommunicating.
* Lars Marius Garshol
| I don't see how that could work, given that the label/identifier
| distinction must be maintained.
* Jan Algermissen
| I think, that the _SAM_ will just provide the assertion patterns
| that are needed to express the semantics reuquired by the SAM.

Well, let's be clear: the SAM is an infoset-based model.  What you are
talking about is the SAM-RM mapping, or representing the SAM in RM
terms. Other than that, I agree.

* Lars Marius Garshol
| I think you are being lead astray by the surface syntax here. What
| the proposed syntax is *really* doing is to distinguish between two
| different kinds of names: labels and identifiers. These have
| different semantics
* Jan Algermissen
| Yes, but is it inherent to the concept of 'name' that it acts as an
| identifier or as a name ? To me these semantics are in the
| _relationship_ between a name and a topic. 

They definitely are (but note that you are being very vague, so this
is eay to agree with), and the proposed resolution to the
topic-naming-constraint issue is that it can be either a label or an
identifier (but that *both* are names), and that the author is allowed
to choose for each individual name which of the types it is supposed
to be.

| Is there any theoretical background that supports either Lars' or my
| point of view, can anyone help ???

What do you mean?
* Lars Marius Garshol
| [...] and cause different processing. They are *not* the same kind
| of name at all.
* Jan Algermissen
| So, if we are talking about how to process them, why are we not
| talking about processing models ;-) ?

Because the term processing model isn't in use any more.
* Lars Marius Garshol
| There are no official plans for an XTM -> RM processing model. What
| the roadmap calls for is an XTM -> SAM deserialization
| specification[2] and a SAM -> RM mapping (as yet unpublished and
| presumably unwritten).
* Jan Algermissen
| Ok, I understand what you mean, it is certainly more precise,
| because XTM -> RM implies that the application MUST implement the
| RM, sorry if I appeared to be proposing that.

Well, actually it doesn't apply that. The committee hasn't yet
discussed this, but what I am currently thinking is that there will be
a separate conformance clause for each part of the new ISO 13250
(which I think ISO requires in any case) and that they may look
something like this:

  - SAM: applications conform to this part of the standard if they
    a) represent all the information in this model and b) document the
    relationship between their representation and the model,

  - XTM: applications conform to this part of the standard if they can
    import and export XTM documents and a) distinguish between valid
    and invalid XTM documents and b) round-trip from a conforming XTM
    document to a logically equivalent XTM document. (The standard
    will of course document what logically equivalent is[1].)

That way, you can conform to the XTM part *without* necessarily
implementing the SAM directly. Similarly, the TMQL conformance clause
might be phrased in such a way that if you can do

  XTM file + TMQL query -> correct result

you are a conforming implementation of the TMQL standard, and although
the TMQL standard is specified in terms of the SAM it's nobody's
business how you actually implement it so long as you produce the
desired result.

Of course, if we ever include a standardized topic map API as part of
ISO 13250 or as a separate standard things will change, but this is
what my somewhat undeveloped thinking has been so far.
| I don't see the RM being an issue in our discussion and I felt that
| the way I said some things might cause that impression.

I understand now. I wasn't confused but you are certainly right that
that bit was less clear than it might have been.

[1] <URL: http://www.ontopia.net/omnigator/models/topic_complete.jsp?tm=tm-standards.xtm&id=op-topic-map-compare >

Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
ISO SC34/WG3, OASIS GeoLang TC        <URL: http://www.garshol.priv.no >