From: Russell Kliese (russell_at_[hidden])
Date: 2005-09-06 21:24:11
> > I am really surprised to see everyone agreeing that a header-only
> > implementation is preferable.
> I completely agree with that sentiment.
I also agree. Header only libraries are convenient, but they don't scale
well for the reasons that have already been brought up. Compile time is
one issue. Also, as the library gets bigger, more and more code will be
compiled into each binary that could have rather been precompiled and
shared from a shared library. Most people will probably be linking
against at least libstdc++ anyway, so it's no big deal to link against
another library (although I only have experience with the GNU tools ---
maybe the Microsoft tools are a different story).
Good to see development on the threads library. In particular, the timed
mutexes now seem to be passing some tests I have been doing.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk