Generic algorithms, was Re: dlw_int/dlw_gcd review summary

From: helmut.zeisel_at_[hidden]
Date: 2001-10-10 02:37:35

--- In boost_at_y..., Daryle Walker <darylew_at_m...> wrote:
> on 10/9/01 6:33 AM, Helmut at helmut.zeisel_at_a... wrote:
> 

> >>> --------------- dlw_gcd -------------
> [SNIP]
> > The more severe restriction, IMHO, is that you
> > require an absolute value of the same type as your ring elements,
> > which indeed restricts your algorithm to integer-like classes.
> 
> The original versions weren't guaranteed for anything beyond 
integer-like
> classes anyway.
> 

I know.

I wrote "IMHO" because I really want to point out
that this is my personal opionion.
As I deduce from the other (non)reactions,
other people do not share this opinion.

Essentialy what I want to say is less specific
to GCD but more general:
When writing a generic algorithm it is important
to find the correct domain where the algorithm should work.

Clearly this involves a trade-off between the danger
that the algorithm promises to work for a domain where
it has not been tested properly 
and the danger that one restricts the domain
too much by adding unnecessary restrictions.

Finding the correct domain can be difficult;
and as the example of GCD shows,
opinions whether the correct domain has really been found
may differ.

Helmut