[std-interval] C++ committee feedback

Steve Clamage Stephen.Clamage at Sun.COM
Fri Apr 7 04:03:18 PDT 2006


Your summary looks right to me. The some points that helped persuade the group 
to move forward with intervals were these:

1. The domain experts (you guys) will reach consensus on what the library should 
look like by October. Except for Sylvain, nobody on the committe has the 
expertise to design the library or evaluate competing proposals. Without 
consensus, the project is doomed. So let's not get caught up in too much trivia. 
The goal should be a library interface that you all can live with.

2. Intervals are not just an academic exercise. Intervals are used widely in 
important industries to aid in critical designs.

3. Interval implementations exist and get a lot of use. The proposed library 
will be based on experience.

4. Although implementation is difficult, a reference implementation of the basic 
functions (arithmetic and transcendental functions) is freely available as Open 
Source from Sun.

---
Steve Clamage, stephen.clamage at sun.com

Sylvain Pion wrote:

> Dear all,
> 
> I just raised the issue in front of the Library Working Group
> of the ISO C++ committee, with the help of Steve.
> 
> The feedback is positive and encouraging.  A straw poll has been
> taken, on whether we should continue the effort :
> Strongly in favor = 4
> Weakly   in favor = 5
> Weakly   against  = 0
> Strongly against  = 0
> 
> The question was also raised whether we should target C++0x or TR2,
> the preference is for TR2 (10 votes against 1).
> 
> [
>   TR2 is a non-normative document which serves as explicit exposure of
>   the intent of the committee for the next standard.
>   TRs are usually also implemented by compiler vendors and serve to make
>   sure users are really happy with the specification before the final
>   blessing.
>   C++0x is the code-name for the next standard.
> ]
> 
> 
> The goal is now to reach a consensus here and have a revision of the
> proposal for the next meeting, beginning of october 2006.
> 
> 
> We also discussed 2 issues briefly:
> - pass-by-value versus pass-by-ref has already been debated by the
>   committee for std::complex.  Seems like the debate was as hot as
>   it is for intervals.  That's an issue for the committee.
> - there are still concerns on the exact list of functions that we
>   will propose, we need to make sure we have the good set (sufficiently
>   useful, specified with enough details such as the width of intervals,
>   and taking into account the difficulty of implementation).
> 
> 
> Steve, please correct me if I did not get something correctly or if you
> want to add something.
> 




More information about the Std-interval mailing list