[boost] Re: Canonical form for Type Traits specializations?

From: Markus Werle (numerical.simulation_at_[hidden])
Date: 2003-11-28 04:58:12

Eric Friedman wrote:

> Maybe a different way to understand my argument: why isn't MPL mentioned
> even *once* in the Type Traits documentation?

Probably because mpl is new and only few have recognized 
that using boost::mpl will be the canonical way for any 
compile time calculation, like using STL is it for for the runtime part ...

> Maybe it's because no one has gotten around to updating the docs. But I
> believe its because the two are not and should not be related on a
> conceptual level.

I recently thought about the library coupling problem 
in boost, because it has become so large and you often
hear people say that they only need this and that and
hate to install 47Megs + binaries at their customers machine.
(Btw: has boost::build a partial-build feature?)

I came to the conclusion that the coupling between 
boost libs is not a problem, since boost is not large 
but small: I "live" on a 5 GB Linux world and some extra
lib like boost makes no difference, especially when looking
at actual hard disk prices and compared to KDE or IE
or jboss.

You could define different levels of - well how should 
I call it? - coreness in boost:
Range goes from some bare metal basic feature like type_traits 
to elaborated end user software like boost::spirit and 
boost::python, but again I feel like such a distiction
is often difficult. It is all designed for reuse.

A tight coupling of the core libraries should 
always happen.
Since boost::mpl is designed as a core library, we can
conclude that type traits can (and should)  reuse it 
and vice versa.

Markus

-- 
Build your own Expression Template Library with Daixtrose!
Visit http://daixtrose.sourceforge.net/