[topicmapmail] tagging vs topics

[Lars Marius Garshol]

* Tony Coates
| 
| I think there is a problem of differing perspectives here.  The
| existing topic map cognoscenti have viewed XTM as a transfer syntax
| for use between topic map engines.  So, the question of integrating
| XTM and document XML never arose, since topic map engines don't
| store documents.

It's been clear all along that XTM and XML would have to be
integrated, but it has also been clear that there are lots of ways of
doing it, and that being able to include XML content as <resourceData>
in XTM documents is not necessarily the ideal solution.

It's true as you say that topic map engines don't store documents, but
they *do* store references to them, and they do store text fragments.
So this is definitely manageable with XTM as it is (see my reply to
Jim Mason).

| Indeed, XTM doesn't even allow non-topic-map XML fragments, on the
| basis that the topic map paradigm should be used to encode all
| information and relationships.

Actually, this has nothing to do with XTM as a format. The decision
was taken to make the topic map data model completely different from
the XML data model (which makes perfect sense). This means that if you
want to say something in a topic map, you have to do it in topic map
terms. (This is identical to what RDF does.)

That does *not* mean that you have to represent your assertions in
a particular way, however. For the past several months I've been
working on Ontopia's Topic Map Autogeneration Toolkit, which can
generate topic map data from any kind of input.

I've used this system to process lots of different kinds of data from
public sources, and found it very easy to integrate topic map
information with information expressed as XML. Basically, you can lift
statements out of their XML form and automatically turn them into
topic map statements, and you can automatically merge them with
statements already made in topic map form. If you want to have
backlinks to the XML documents that produced the information in the
form of occurrences that's no problem.
 
| The usual story is that you should use the topic map to point into
| your document.  That allow providers of topic map engines to wipe
| their hands of the problem, but does little to alleviate the
| document handling problems of users.  This advice tells users that
| if they use topic maps, they will be forced into multi-file
| solutions, and that may not be a palatable story.

Well, no. My reply to Jim pointed out that there are ways of doing
this with a single XTM file, and that there were ways of doing this
without using files at all.

| It would be useful to compare this with RDF and NewsML.  RDF can
| annotate any general XML document, and does not care that there is
| non-RDF in the document (well, RDF treats all XML as RDF, but that
| is another story).  It certainly allows you to build rich
| information into the document file, and have a 1-file solution.

Topic maps are no different in this regard. True, there is no
standardized way of turning any kind of XML into topic map data, but
there are non-standard ways of doing it.
 
| This is not to say that XTM as is should be changed to allow
| inclusion of general XML.  But it does beg the question as to
| whether there should be an equivalent embeddable markup that allows
| XML documents to carry their own topic map of what is in the
| document, and what external items are related to the document,
| without requiring external files.

You *can* do this; you just have to make a solution for turning the TM
information in your XML document into topic map data. There are lots
of ways of doing that: using XSLT, embedding LTM fragments, writing
your own SAX/DOM application, using a generalized toolkit, etc.

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