Re: OONSTD: to early for standards?

Gabriel Dos Reis (Gabriel.Dos-Reis@dptmaths.ens-cachan.fr)
Thu, 2 Jul 1998 02:54:28 +0200 (MET DST)

>>>>> «Jonah», Jonah Lee <jonah@mealpha.engr.uaf.edu> wrote:

Jonah> If an initial standard has a basic set of functionalities, then new
Jonah> features can be added later.

In general, a standard committee is mandated to standardize common
practices. In our affair it just happens that there are no sound common
pratices -- at least as far as C++ is concerned. There are
experimental techniques we are trying. Personnally, I think we are at
the beginning of a long journey. So let's explore *in depth* the country
before definitly casting in stone whatever we have.

Jonah> This situation probably will always exist, standard or no standard. But
Jonah> having an initial standard would at at least ensure some interface
Jonah> compatibility. For example, I have not used a Fortran compiler that does
Jonah> not have its own extensions.

Nothing prevents someone from developing his own library. What I'm
against is to standardize practices that have not been fully
tested. If we were to add features latter, why not _to develop_ a
library, *fully* test it and *then* standardize it instead ? But
we're far from the beginning.

Jonah> These should not then go into the initial standard and can be added
Jonah> later.

That is not a wise position. This reflects amateurism more than
professionalism.

Jonah> I would suggest that current and future library developers who have a need
Jonah> of such a library and they may want to work together. Many libraries are
Jonah> developed in the scientific community so the incentive is to have better
Jonah> software and be more productive. It may be easier to get people together
Jonah> and reach some agreement when the market is small (maybe) ?

Experience proved me that there no canonical way to
numerics. Libraries developers need not to concentrate on a single
library. I'll rather encourage them to develop their own idea *but*
they should keep in touch, communicate their work.

Jonah> It would be nice if such things exist in the OO world I guess that was why
Jonah> this discussion got started.

I agree. But For the moment there no point to standardize immature
constructs.

Best wishes,

-- Gabriel