Re: [tm-pubsubj-comment] formal syntax (was: Tuesday Conference)

[Murray Altheim <m.altheim@open.ac.uk>]

Bandholtz, Thomas wrote:

>  > -----Original Message-----
>  > From: Murray Altheim [mailto:m.altheim@open.ac.uk]
> ...
>  > You have plenty of criticism of XTM, though given that XTM contains
>  > the ability to establish a fairly rich graph using a simple syntax,
>  > there's absolutely no reason why XTM syntax cannot be used to
>  > describe a rich taxonomy/ontology. It's a matter of developing the
>  > topic and association types used in that graph, and this TC is trying
>  > to describe how these should be published.
> 
> OK, I checked that: I can express anything I want to say in XTM, if it 
> has to be (*sigh*). As I have seen here in Barcelona, XTM is more or 
> less accepted as a worthy contribution by at least some relevant RDF 
> people here ("if you have PSI, why shouldn't we use them").

If one separates the technical from the political/marketing hype, there's
no reason why XTM can't be used for a *lot* of applications. But due to
it being in seeming competition with RDF, it's received a lot of undue
criticism and attention from those whose foundations got a bit less
secure with the advent of XTM.


> I would prefer to have an XML Schema, sorting Typology from Topics, 
> using inheritance, modular XML namespaces, etc, etc. But there are 
> samples pro and con: Open GIS Consortium has started GML using DTD and 
> then moved to Schema "on-the-fly" and now they are pretty stable and happy.


I seriously doubt that had so much to do with choice of schema language.

Complexity in the end will kill them. As comparing Hytime and HTML, nobody

wants complexity, they want simple things that work. Only designers want
complexity because it satisfies their aesthetic sense of importance.


> SVG is a lovely standard, and they use a DTD. Anything goes, if the 
> standard is strong itself.


I don't quite follow the idea that the schema *language* used to define
a markup language is somehow significant here. One could create an
XML Schema, a DTD or RELAX grammar to describe XTM, and indeed I believe
there are transformation tools to transform one to another. There are
advantages and disadvantages of each, but that's another discussion
entirely. The only reason you're seeing XML Schemas used to describe
new W3C markup languages is that TimBL decreed that it was now a
requirement (so much for technology to fit the need). It's all marketing.

As for SVG being a "lovely standard" I won't divulge the ugliness of
the process that created it, nor that it would probably be a lot more
successful if it wasn't designed to match PhotoShop's feature set. I've
advocated modularizing SVG so that features such as animation, filters
effects, etc. could be eliminated from a simpler profile to enable
more companies to play in the field. Adobe wouldn't have it.


> But, tell me, why not use XTM for PSI then? Are we ashamed of XTM?


Due to time and resource constraints I opted to bow out of full
participation in the TC, so I'm not a voting member. That said,
had I a vote, I'd vote against any new syntax, or any publishing
of PSIs in some alternate syntax than XTM. I think XTM is both
suitable and necessary for the publishing of PSIs. PSIs can fit
within an ontological structure (i.e., one PSI can have a specific
relation to another), and I'd like to see those relations expressed
as XTM <association> elements, not some new syntax.


>  > I'm in a sense waiting for something to settle before I publish a
>  > preliminary set of logical primitives as an XTM topic map, which is
>  > then usable in creating taxonomies and ontologies. Currently, XTM
>  > 1.0 includes association types for superclass and subclass (as PSIs),
>  > so taxonomies are already representable. Just as you mention XSLT,
>  > it's certainly possible to transform from any other XML-based
>  > taxonomy syntax into XTM and vice versa. I was in the process of
>  > working on Cyc when the OpenCyc project began, so I'm now waiting
>  > on that one too. *sigh*
> 
> I have to confess I do not know OpenCyc. Tell me more.


Doug Lenat started up the Cyc ontology project many years ago, and
announced at the 1999 or 2000 Markup Technologies conference in
Montreal that he'd be giving away (in stages) the entirety of the
Cyc ontology, starting with the upper ontology. Cyc is a "common
sense" ontology (check cyc.com). Plans changed and it became an
"open source" project called OpenCyc, on SourceForge. Quite
interesting, as it comes with a full Java-based framework now,
complete with query engine.


>  > Last year, Peter Becker and I began working on an XML syntax for
>  > Conceptual Graphs but were unable to figure out the abstract model
>  > behind it. Sowa et al are now working on Common Logic (which will
>  > include an XML serialization syntax), which I believe could be
>  > transformable bidirectionally into XTM syntax. If nobody else does
>  > it, I'll do it. But we'll have to wait awhile for that one, as I
>  > don't that train is moving that quickly right now.
> 
> If so, we should consider Schema !!! One more argument: we will need it 
> to play a role within Web Services and ebXML.


No we won't. Maybe you will. I'm going to watch "Web Services" pass like
every other marketing bubble. Schema doesn't give us anything we don't
already have, as we're not trying to devise a constraint language for
documents, we're trying to design a means of publishing PSIs. XML Schema
would only add a layer of useless complexity to this project.


>  > I really don't see all the weaknesses you describe as being barriers
>  > to implementation. No syntax is ever perfect but I believe XTM is
>  > sufficient, that we hit the 80/20 point pretty dead on. Those in
>  > the "20 league" will always want those missing features, but adding
>  > them or abandoning XTM at this point would simply cause people to
>  > abandon topic maps. We need some stability more than we need
>  > perfection right now, both in terms of syntax and specification.
>  > This TC's goal is to provide a simple means of publishing Published
>  > Subjects. If the TC fails, it will fail due to making the methodo-
>  > logy either too difficult or too removed from XTM, in my opinion.
>
> OK, then let's use XTM for PSI, or drop XTM completely (no joke)
> We might even use RDF for PSI. But - *please!!!* not a third thing.
> If I can find the time, I would even think about re-writing XTM using 
> XML Schema, so that any valid XTM document would remain valid.
> 
> I also had discussions with Steven Newcomb and Michel Biezunski today 
> about adding an event-type with a temporal extent attribute and a 
> location-type with a gml:boundingBox. May this points to the direction 
> of an XTM 1.1 ....


I don't think we need an XTM 1.1 to do that. You could design event
types, temporal modes, location types, etc. using the existing syntax
and PSIs. It's possible that with a set of PSIs for say, predicate
logic, one could create an inferencing engine that could take XTM
as source documents. I keep thinking about a KIF-to-XTM translator,
but I don't think it's worth the trouble. I'll wait for Common Logic
to appear, then give it a go.

Murray

......................................................................
Murray Altheim                  <http://kmi.open.ac.uk/people/murray/>
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK

      In the evening
      The rice leaves in the garden
      Rustle in the autumn wind
      That blows through my reed hut.  -- Minamoto no Tsunenobu