Replacement for GMP: Update

[Thorkil Naur]

Hello,

On Thursday 10 August 2006 07:31, Peter Tanski wrote:
...
> Summary: I finally settled on modifying OpenSSL, since that would be  
...

Being a heavy user of Haskell Integers, I have followed this development with 
great interest. Although your decision has its drawbacks, it could very well 
be the best, all things considered.

I would like to mention a few things that I have not seen discussed: Clearly, 
using an existing library unmodified would be preferable: New developments, 
error corrections, documentation, wide exposure, all of these things would be 
available.

I have looked briefly at the OpenSSL Bignum library and in the areas of memory 
management, but also error handling, it seems clearly intertwined to some 
extent with OpenSSL in ways which would appear to rule out the direct use of 
this library for GHC Integers. However, considering the advantages of using 
an existing library unchanged, we might consider another possibility: Working 
with the OpenSSL people to modify their library to allow the sort of 
interfacing that is needed for its direct and efficient use in GHC. While, of 
course, retaining its value as part of OpenSSL.

(And way further back: Have we tried to discuss the LGPL licence of GMP with 
the authors? I am not really into all these matters, sorry if this doesn't 
make sense.)

Failing that, I would suggest considering the development of the modified 
library to a form that would allow independent use, apart from its use in 
GHC. This would add valuable possibilities to your options when choosing the 
precise mixture of Haskell and, perhaps, raw C code that best balances your 
performance desires and needs for convenience.

I wish you the best of luck with your work.

Regards
Thorkil