[topicmapmail] Querying Topic Maps

[Martin Bryan]

Murray
>
> This argument is based on the erroneous assumption that common XML-based
> query mechanisms can be used to query either ISO 13250 or XTM topic map
> documents. As was pointed out to me several months ago (admittedly to my
> dismay), these solutions simply cannot work because a topic map document
> (represented in either syntax) is not simply an XML document. An XPointer
> or XPath query would ignore proper merging and provide an erroneous
result.

Note that I've changed the title on this one, as I believe it is the
starting point for a serious philisophical discussion ;-)

The question arises as to whether or not it is valid to merge topic maps
simply based on the commonality of names and/or public subjects. Whilst I
can see the validity of merging on public subjects, I have problems with
name-based merging, and even with public subjects I personally  believe that
no topic maps should be merged without speficying added themes that can be
used to differentiate the occurrences and associations derived from each of
the maps. (This is especially important for XTM, where there are no other
ways of distinguishing the source.)

I personally believe that safe merging will only occur when the added themes
or scopes of a pair of entries match. When adding themes you (should) have a
URI for the referenced topic map documents. If you have a set of URIs then
any XML query can be sent to the set of URIs, so that you can then identify
entries from sets of maps using common XML-based query mechanisms. I simply
do not believe that XPointer/XPath need to search the result of the mapping.
For topic maps to interwork on a web-wide scale, of the type postulated for
the Semantic Web, we need to be able to query maps in multiple sources
without having to copy them into a single computer's memory.

Martin