[topicmapmail] Not just Temporality in Topic Maps, but a whole approach

[Simon Grant]

At 19:26 2006-06-12, Bernard Vatant wrote:
>who Simon Grant was, and guess I found out at [1].

Indeed - sorry I didn't make it clear earlier - I often sign my 
e-mails with my home page address.

For the rest, I note and respect your position. I am still inclined 
to differ. Let me try to express a bit of the reason(s).

>[...] Now if the question is : are topic map - based systems more 
>likely (than other systems) to meet human requirements in terms of 
>interaction, I think this question is a non-starter.

I would unequivocally agree with the position that in principle, TMs 
are no more *able* to represent user thinking or *able* to meet 
requirements than any other technology. Undoubtedly one *can* do all 
kinds of transformations and swaps. That is not my argument. But let 
me continue that same line - there is no reason in principle why 
knowledge should not equally be represented, and requirements met, in 
lisp, in Prolog, or even in Java. As you imply, these are just technologies.

But I wouldn't suggest that you believed it was equally effective to 
represent knowledge in Topic Maps or in C. Once significant 
differences have been admitted between different technical approaches 
and different formal structures, the way is open (indeed it cannot be 
closed) for asking which of the various technologies may be more 
effective in different ways than other ones.

I can understand why people might think that the end user 
understandability of the representation scheme doesn't matter. 
However I disagree, carefully. A short (and perhaps inadequate) 
representation of my view is that the more the implementation 
diverges from the user's own language and concepts, the more work 
will be needed by the implementers, the more time and money it will 
take, therefore the less work is likely to be done within real world 
resource constraints. Also, the wider the gap, the more room there is 
for misunderstandings to grow.

>In Mondeca we've been working with the same generic data model for 
>six years. We've implemented integrated systems around it "as if" it 
>was topic maps, then "as if" it was RDF, and "as if" it was 
>conceptual graphs, and what else tomorrow, without changing anything 
>at the core. Some implementations/user interact as if it were a 
>regular data base with capacity for "smart" queries, other a 
>repository of reference terminology for text mining tools, other a 
>knowledge base, other a thesaurus management tool, you name it. The 
>key is genericity, flexibility, and imagination in implementation. 
>No free lunch :-) .

What people seem to mean when they talk about the lack of free lunch 
is that there are no solutions which solve everything without any 
effort. That much is clear to most rational people. What people seem 
to be in danger of forgetting is that some approaches are more 
effective than others. Saying that there is no free lunch risks 
trying to obscure the fact that there are differences, and in some 
cases (I wouldn't like to suggest for a moment that anyone on this 
list does this) could be imagined to justify not comparing the 
alternatives, but instead sticking to the approach with which one is familiar.

In the case quoted, I can well imagine that if the underlying 
technology is the same, there would be little difference in making 
the application behave as if it were TM, RDF, or whatever. Again, 
that is not the point.

Simon