[boost] Re: math_constants parameterisation and return by ref

From: Ken Hagan (K.Hagan_at_[hidden])
Date: 2004-01-23 05:42:21

> Paul A. Bristow wrote:
>
>> There was widespread agreement that no solution requiring pi() would
>> be widely acceptable.

Deane Yang wrote:
>
> I'm sorry if my recollection is incorrect, but I only remember seeing
> messages from you stating this. I do not remember seeing anyone else
> on the boost mailing list expressing this view.

As I recall, someone from Fermilab claimed that they had tried
the pi() syntax, but the end-users (your average scientific
programmer) just wouldn't use a constant that looked like it had
function call overheads. Since this was real-world experience
with the intended user base, it carried quite a lot of weight.

It is also my recollection that most boosters who expressed an
opinion were somewhat shocked that a group of degree-qualified
scientists couldn't see (or just accept) that there was no such
overheads. Perhaps that was just my reaction. :)