[sc34wg3] RE: Topic Maps Reference Model posted (new draft)

[Patrick Durusau]

Bernard,

Apologies for the very late response to your comments! I can post to the 
list but apparently I am not getting messages from the list?

Checking on it now but wanted to go ahead and respond to your post.

> Patrick and all
> 
> Some first comments on the new TMRM
> 
> First point that strikes me is that main (only?) focus is now really on *subject
> identification process* - unless I have a strong personal bias towards seeing my own
> obsessions everywhere. A TMA appears to me as nothing more than a context in which a
> certain set of *subject identification rules* hold - which should be made explicit by the
> TMA in order to be interoperable with other contexts.
> The document still speaks, though, but IMO more out of an old habit than anything else,
> about *subject identity* as something that could, at least in theory, be defined in some
> absolute way. But actually it is mainly interested in *ways of identifying subjects* by
> comparing their proxies' properties in a given TMA context, or across several contexts.
> Accordingly, it implicitly admits that definition of any absolute identification process,
> IOW definition of absolute subject identity outside the context of any TMA, is neither
> feasible, nor even desirable. Very much agreed!

Yes, quite correct.

Curious what part of the document you see as saying that "...*subject 
identity* as something that could, at least in theory, be defined in 
some absolute way."

We attempted to address the notion of subject identity only within the 
context of a TMA. We were simply silent about subject identity outside 
of the context of a TMA.

Whether absolute subject identity is possible, desireable, etc, within 
or without a TMA is not a question we needed to address and so simply 
said (or tried to say) nothing about that issue. A TMA defines what can 
be said to identify a subject and that was our only concern.

> My hunch is that the next step will be to get rid altogether of the notion of (absolute)
> subject identity, the same way in the last century Quantum Mechanics has get rid of the
> notion of (absolute) real-world object. The same way a quantic object only appears (IOW
> "comes to existence") through some interaction with the observer, a subject always appears
> in some process of conversation between various actors - humans and/or computers. And the
> same way non-ambiguous interpretation of a quantic experiment is possible through a
> well-defined explicit experimentation protocol, non-ambiguous subject identification needs
> a well-defined explicit processing context (call it TMA if you like).

+1

As I said, I don't see 'the notion of "absolute" subject identity' in 
the current draft but then I am far too close to it. Suggestions for 
revision?

> All in all, it goes in the direction a bunch of people have been trying to plough for
> quite a while now in various places, and that I keep tracking on universimmedia blog.
> Strangely enough, all the references in Annex A had been posted lately there ... such a
> convergence makes me figure that something is in the air. Either we are all go wrong
> together, or it's really the way ahead now.
> 
> Second point is that disclosure of identification rules is certainly a necessary condition
> to allow interoperability across various TMA, but I'm convinced it will never be
> sufficient to ensure "universal" interoperability.
> 

Well, but the goal was not "interoperability" but "integration." The 
difference being that XML, for example, assuming it is valid, etc., 
gives us "interoperability" but no guarantee that the resulting 
information is meaningful.

Knowing the identification rules *enables* integration across various 
TMAs, but certainly there is no guarantee that is going to be successful 
in any particular case.

I think it can be guaranteed that without knowing the identification 
rules, that no automated integration is possible. (You could write the 
rules ad hoc for any information you wanted to merge together but then 
you would have the rules that I posit as necessary for integration.)

> Quoting the introduction
> 
> "... indications of subjects are recognized only within limited ontological, cultural
> and/or technological contexts. If we want to gather all the information about a given
> subject but the ways in which the subject is indicated are ontologically, culturally,
> and/or technologically diverse, our task is only possible if we know what those ways are,
> and how they work."
> 
> I agree with the first sentence, but would happily restrain the second one to acknowledge
> that all we can expect is to provide ways to extend as far as possible contexts of
> identification, across various TMA using different identification rules, providing those
> rules are not as different as even their sheer comparison becomes impossible. There is a
> breaking point where comparison of rules would need either some universal rule language,
> or the recursive exercise of identification of concepts used by the rules exposed ...
> 

Leaving aside the point that the introduction is non-normative, isn't 
the discovery that two sets of rules are so different that comparison is 
impossible one that is covered by "...our task is only possible..."? We 
tried to be careful not to promise successful comparison, merely to 
enable it.

> IOW, it would be sensible for the document to acknowledge the limits of the process it
> defines, and the relativity of its central concept, subject identity (switching to subject
> identification). That would not make the RM less universal. On the opposite, to
> acknowledge identification relativity, but providing tools to arbitrarily extend the
> context in which it can be applied, is in fact more powerful that pretend to grasp
> universal identity.
> 

Curious what you see as the "limits of the process it defines...." By 
definition, any system for identifying a subject has limitations, I 
think we agree on that, as do processes for disclosing them, I think we 
agree there as well.

But the Topic Maps Reference Model does not define any system for 
identifying subjects nor does it define any processes for disclosing 
such a system. It sets the requirements of disclosure, not how it is 
accomplished.

Are you suggesting that we say that systems for identifying subjects, 
the systems for disclosing those systems as well as subject identity are 
limited and meaningful only from a point of view? I have no problem with 
that but it looks like gratuituous advice on an issue we don't have to 
reach.

Hope you are at the start of a great week!

Patrick

-- 
Patrick Durusau
Director of Research and Development
Society of Biblical Literature
Patrick.Durusau@sbl-site.org
Chair, V1 - Text Processing: Office and Publishing Systems Interface
Co-Editor, ISO 13250, Topic Maps -- Reference Model

Topic Maps: Human, not artificial, intelligence at work!