Boost logo

Boost :

From: Douglas Gregor (gregod_at_[hidden])
Date: 2002-09-06 11:22:54


On Friday 06 September 2002 11:31 am, David Bergman wrote:
> If we get that formal, we could not apply "equivalence", or virtually
> any of the terms we have used throughout this thread...

Precisely.

> I think we are better off with translating the arguments into the
> classical "true or false" world, in order to have nomenclature at all.

If the nomenclature doesn't apply in the 3-state world, why are decisions made
using that nomenclature applied to the 3-state world? This logic just doesn't
follow: Narrow the 3-state definition to 2 states; there isn't any proper
equivalence relation, so the 3-state definition is not a good one.

I feel that this thread has gone way off-topic for this library. When
designing a library, we are faced with the task of finding the best interface
to suit the set of tasks the library is to be used for. The problem is
two-fold: (1) we need to determine the set of tasks that can be solved with
the library and (2) we need to construct an interface that fits well within
the domains of these tasks.

However, in this thread we have wandered off into the question "What is a good
semantics for interval relations?", which I would argue is irrelevant to
library design. The related, relevant question we _should_ be answering is
"What semantics for interval relations are needed in the application domains
we want to support?"

        Doug


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk