On 16 April 2010 17:02, Sergey Bochkanov <[email protected]> wrote:
> Hello, Bill.
>
> You wrote 16 апреля 2010 г., 15:07:18:
>
>> The things wrong with LGPL:
>> * the number of contributors is too small even after two years of
>> working on this project
>> * companies are suspicious of it - worried the license will be changed
>> to GPL (I know this for a fact, I'm not surmising)
>> * if we use version 2.1 we can't use code from GMP and contributors
>> who refuse to use v3+ only (of which there are numerous)
>> It's BSD or v3+. We don't have any other viable options. And we've
>> demonstrated that. If MPIR was going to get piles of regular
>> contributors, we would have done by now.
>
> Thanks for the clarification. One more question: what's with interface
> of the new library? Will it be close to the GMP's one? That would make
> porting existing applications much more easier.
>
> Of   course,  there  are things  that are just source of confusion. It
> would be good to drop them from new library. For example:
> * inconsistent sizes of mp_exp_t/mp_bitcnt_t/... Sometimes they are 4,
>  sometimes  they  are 8 (depending on compiler, OS and CPU). It makes
>  writing cross-platform programs harder. Using GMP  from  another
>  languages  becomes harder too.  Many  modern  languages  aim  to  be
>  platform-agnostic, i.e. program should work  independently  of  the
>  32/64  bit  issues.  But it is hard to achieve when  underlying  GPM
>  API  changes  when  moving  to  another  platform.  C#  program, for
>  example,  works  as  32-bit  under x86, but it works as 64-bit under
>  x86_64.

This is, in my opinion, difficult to do anything about. The efficiency
of arithmetic heavily depends on whether you use 32 or 64 bit, and GMP
and MPIR aim to be efficient.

However, I also believe it is becoming less important. Most Windows
users have finally heard that 64 bit machines exist and some of them
are actually buying them. So soon, everyone will have a 64 bit machine
and 32 bit machines will not exist.

In my opinion, 128 bit machines will never become useful for mp
arithmetic, if they are ever built. So the situation seems to be
normalising somewhat (other issues are becoming more important, such
as parallelism).

> * lack   of  normal  error  handling (out of memory, division by zero,
>  mpf_t underflow)

I'm not sufficiently expert at floating point arithmetic to comment on
this. The experts tell me that the mpf stuff in MPIR could be dropped
without anyone caring. That suits me fine. I think MPFR should be used
for such things as it is much better.

As for future directions of the library. I think it is too early to
set any long term objectives. The key to success will be to get
something up and running as soon as possible, and so this means
limiting the goals. For my money, that means 64 bit only, from the
start, and concentrate on implementing the GMP mpn level only (albeit
renamed). In 6 to 12 months we can re-evaluate and see how successful
we've been.

I've had more off list support for a BSD library today. See that never
happened for MPIR. I had completely underestimated how much support
there would be for a BSD licensed library.

We'll see how much the enthusiasm turns into code. That is something
we will have to evaluate over time.

The next steps will be to get a website and devel list up and running.
The latter is trivial and I will do it as soon as we have a website
up. Watch this space....

Bill.

>
> --
> With best regards,
>  Sergey                          mailto:[email protected]
>
> --
> You received this message because you are subscribed to the Google Groups 
> "mpir-devel" group.
> To post to this group, send email to [email protected].
> To unsubscribe from this group, send email to 
> [email protected].
> For more options, visit this group at 
> http://groups.google.com/group/mpir-devel?hl=en.
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"mpir-devel" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/mpir-devel?hl=en.

Reply via email to