The GNU Scientific Software Library

[E. Robert Tisdale]

Brian Gough wrote:

> GNUSSL is a C++ class library, covering areas of linear algebra,
> matrix decompositions and FFTs in an object oriented way.
> So it is fairly different from GSL,
> where the goal is to write a portable replacement in C
> for existing numerical libraries
> such as IMSL/NAG/Numerical Recipes.

But the GSL does NOT implement
the IMSL, NAG or Numerical Recipes APIs.
You could just as well argue that
the GNUSSL was a portable replacement
for existing numerical libraries.

What distinguishes the GNUSSL from the GSL?
What is the compelling distinction
that helps numerical application programmers
choose one library over the other?
It the GSL always the better choice?
Or are there circumstances where the GNUSSL
would be the better choice?

> The projects began around the same time (roughly speaking)
> but are independent.

Why is this?
Was there nothing that the GSL developers could learn
from the GNUSSL developers?
Was there any attempt to make the two libraries compatible?
Do they complement each other?
Is one of them redundant?

> I believe development of GNUSSL lapsed after a while.
> I'm not sure why that was.

What do you mean by lapsed?
Is the GNUSSL unfinished?

Will development of the GSL lapse after a while?
Do application programmers have any assurance
that the GSL will continue to be supported and maintained?