[tmql-wg] Toma - Use Cases Solution

Robert Barta rho@bigpond.net.au
Sun, 14 Mar 2004 17:15:20 +1000

On Thu, Mar 11, 2004 at 03:36:54PM +0100, Lars Marius Garshol wrote:
> | I found that the use cases are excellent - they really challenged
> | the languages. However I think they don't cover a very important
> | domain which is "introspective queries" (as it is called in the
> | 'tolog for TMQL' presentation). 
> That's actually in section 6, but I don't feel those queries have been
> specified in enough detail yet. I've suggested to Robert that we
> tighten them up, but haven't heard back yet (I guess he's busy with
> something).

Sort of.

The 'needle-pin' queries in section 6 are collected from earlier
efforts. I had maintained them in a TM (like the other use cases).

Most feedback concerning this section was neutral to negative,
actually.  Still I thought it is worth keeping them as they may occur
quite frequently.

If the feeling now is that it worth to work on this section then
(a) Lars and I (I guess I can speak for him here) are happy to
take input or (b) can certainly put some effort into these.

[ Digression:

>From a architectural point of view, one might be sceptical whether
TMQL should really fully support these queries, in the sense "simple
query = short query":

The whole idea about SQL and XQuery, for instance, is to keep the
application "as long as possible in the querying processor". The
reason for this is that long, complex queries are much more effective
to optimize compared to many small queries fired off from the

I have seen RDBs running on quadruple Pentiums on load 20, sustained,
just because some programmers (PHP, what else) tortured the database
with trivialities.


> | Beside that, in my opinion the topic map that is used for the use
> | cases is not the perfect one for such objective. I think that we
> | should use a topic map that is about a domain we all know about, and
> | yet is not too small and simple.
> I don't know. Personally I don't think the topic map used matters at
> all so long as it lets us do all the queries we need to. This is about
> the queries much more than the query results, I'd say. :)

I'd second that, regarding the query use cases maps as rather
isomorphic. Whatever we choose, some of the queries (I spent some time
on this) will sound completely artificial... :-)

> | Maybe it is better also that the topic map is created in XTM which
> | is the standard now (although I prefer for example AsTMa= to
> | describe topic maps manually). 
> We can provide it in XTM, and I think that's all that's needed. I
> don't think how it was created is very important, really.

Well, Rani is correct in the sense, that in


most of these links are broken. In the original, the map itself
is provided in XTM as well as in AsTMa?

I have put it onto my todo list when upgrading the TMQL homepage.