This is the mail archive of the gsl-discuss@sources.redhat.com mailing list for the GSL project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: GSL API design flaws


"E. Robert Tisdale" wrote:
> 
> Brian Gough wrote:
>
> > The goal of the project is really just to provide a free library
> > equivalent to existing proprietary libraries
> > such as IMSL, NAG and Numerical Recipes.
> > So we just follow the existing conventions
> > from these libraries with some minor improvements.
> 
> That's a disappointment.

Yes, it is. But we're not the source of the
disappointment; it is a simple fact that the
state of the art "products" in this field
fall far short of technical expectations.

The problem is that attempts to raise the bar
have never reached maturity. Historically, there
are only two results of attempts to raise the bar.
First are those projects which are expressed as
specialized research projects (such as in the area
of linear algebra) which never attain the level of
robustness, completeness, or support needed to keep
them in circulation. Even worse is the second class
of projects which fail without ever producing anything
useful. Typically this happens when the implementors
attempt to solve all the problems at once, and become
hopelessly mired in their own object goo.

It is possible that the latest round of projects
will reach new levels of maturity
(witness MTL). But the jury is still out.

After all, do not forget that it is only very recently
that any progress at all has been made on the problem
of vector/matrix representation and associated linear
algebra. That took years of effort and several
breakthroughs relating to expression of generic
programming with C++ templates. And that's only
the start of the road. Think of all the problems
that remain, relating to issues such as cross-langauge
development, expression of parallelism (for both shared
and distributed architectures), generic programming
expression for other algorithmic areas, etc.

GSL is not the place for free jazz explorations
on numerical library themes. It's just a hammer.
You use it to pound nails. Higher levels of
sophistication must wait. I don't like it any
more than you do, but those are the facts.


-- 
G. Jungman

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]