Re: [gentoo-science] aldor

["M. Edward (Ed) Borasky" <znmeb@cesmail.net>]

Andrey G. Grozin wrote:
> Hello *,
> 
> I've just committed aldor-1.1.0 to the science overlay. This is the 
> first open-source release (free for non-commercial use). Previous 
> releases were binary-only (free of charge for non-commercial use, x86 
> only). Now aldor and its libraries are compiled from sources in this 
> ebuild, and can (probably) be compiled for any arch.
> 
> Aldor is a language developed for many years at IBM, then NAG, and now 
> aldor.org. It was designed as a replacement of the Axiom library 
> compiler. It is an excellent library for implementing computer algebra, 
> and it has a rather comprehensive foundation library (libalgebra) with 
> many algebraic domains (it is not as comprehensive as the library of 
> Axiom, written in the old language). Aldor can be used both separately 
> and from Axiom. I am sure that the next axiom ebuild must have the local 
> USE flag "aldor" which would allow using aldor from axiom (and will 
> depend on the aldor ebuild I've committed now). What I am not sure is 
> which axiom to package, there are now 3 of them :-(
> 
> I know that the current ebuild is far from ideal. It ignores the user's 
> CFLAGS and uses the ones supplied by the aldor folk. The build system is 
> rather non-standard, and fixing this will require patching a number of 
> makefiles. But in any case, it is better than the previous 
> free-of-charge binaries - they were compiled with the same CFLAGS chosen 
> by aldor developers, and on x86 only. So, please, try aldor-1.1.0 and 
> report yoor experiences. This is a very well designed language for doing 
> mathematics.
> 
> Andrey

I think I've mentioned on this list before that

a. The forking of Axiom three ways was (and still is) a contentious 
ego-ridden mess, and

b. While the Axiom project folks begged NAG to "free Aldor", the best 
they could get was a "free for non-commercial use" license, which makes 
it incompatible with most "free as in freedom" licenses.

So ... while Aldor does indeed appear to be a product of high quality, 
you do have to be careful and choose wisely how you spend your time with 
Aldor, Axiom, OpenAxiom and FriCAS. For now, I am sticking with the main 
branch of Axiom and not touching Aldor until "things settle down a 
little bit."

I'm hoping that the disputes causing the forks will get resolved and 
there will only be one Axiom. And I'm hoping NAG can be persuaded to 
eliminate the non-commercial clause from the Aldor license.

IIRC I posted a bug in Bugzilla to get Axiom up to date long before the 
forks. I'm personally running the main line Axiom on both AMD64 and x86 
with only one issue -- the "gcl" it carries only works in older versions 
on AMD64. If you follow the Lisp mailing list, "gcl" is close to getting 
kicked out of Portage because there are better Common Lisps available. 
But these are issues that I think can easily be fixed upstream without a 
fork. :)

I have tested the forks ... they don't currently offer me anything the 
main line Axiom has. It looks like the *math* is getting maintained in 
the main line and the FriCAS and OpenAxiom forks are more about build 
improvements. Build improvements are sorely needed, but the *reason* for 
having Axiom is to do math, not to rebuild Axiom faster. :)
-- 
gentoo-science@gentoo.org mailing list