|
Boost : |
From: Beman Dawes (beman_at_[hidden])
Date: 1998-12-27 09:46:49
At 08:22 AM 12/21/98 -0800, Jerry Schwarz wrote:
>I haven't looked at the details of the classes myself so I haven't
been
>commenting. But judging from the facts as I understand them from
the
>discussion I strongly favor Greg's idea that there should not be
>parameterization.
>
>A. It is hard to predict what is going to happen to code at the
micro level
>with today's machines and optimizers. It may be that the code space
that is
>lost to having to do an extra indirection is made back from the
advantage
>of having a smaller struct to begin with.
>
>B. Any parameterization will discourage users. It may be that it
doesn't
>matter which variation they pick in most cases, but they won't know
that.
Your points are well taken.
Perhaps the best approach for now is simply to mention some of the
other approaches as a sidebar in the documentation. If there is
later demand for actual implementations, they can be provided without
breaking any existing code, either by parameterization or as
separately named classes.
--Beman
------------------------------------------------------------------------
E-group home: http://www.eGroups.com/list/boost
Free Web-based e-mail groups by eGroups.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk