[boost] Re: Big Int

From: David Abrahams (dave_at_[hidden])
Date: 2005-03-08 15:52:00

Christoph Ludwig <cludwig_at_[hidden]> writes:

> On Tue, Mar 08, 2005 at 10:46:21AM -0500, David Abrahams wrote:
>> "Richard Peters" <r.a.peters_at_[hidden]> writes:
>> 
>> > "Hubert Holin" <Hubert.Holin_at_[hidden]> wrote in message
>> > news:Hubert.Holin-D2560E.18082507032005_at_sea.gmane.org...
>> >> > Well... I try to aim at core operations only. At the moment, I support
>> >> > jacobi symbol and primality testing, but I will almost certainly remove
>> >> > them.
>> >>
>> >>       Huh? Why?
>> >>
>> >>          Hubert Holin
>> >
>> > Because I think they do not belong in a big_integer library. They belong in
>> > a number theory library. If functions like jacobi symbol make it into a
>> > big_integer library, then so should the legendre symbol, lucas function,
>> > rabin-miller, lucas probable prime, etc, etc.
>> 
>> You might want to peruse the discussion between me and Peter Dimov in
>> boost-users:
>> 
>> http://news.gmane.org/find-root.php?message_id=%3cd053gs%249km%241%40sea.gmane.org%3e
>> 
>> Whether these things "belong to a library" at all is open for debate.
>
> I do not get your point. In the thread you referred to the only discussion
> between you and Peter Dimov I see is about about points of customization, ADL
> etc. What's the connection with the jacobi symbol and similar number theoretic
> functions? I do not recall anyone proposing an interface to such functions
> that you are supposed to customize for your_favorite_integer_type. 
>
> Confused...

Look for Peter's arguments about which library "owns" a customization
point.  These functions will apply to a variety of numeric types, and
therefore will be used as customization points.  Someone will define a
concept that depends on them.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com