Re: [boost] Re: Why Jam?

From: Braden McDaniel (braden_at_[hidden])
Date: 2002-02-23 21:22:01

On Sat, 2002-02-23 at 13:14, Peter Dimov wrote:
> From: "Braden McDaniel" <braden_at_[hidden]>
> > On Sat, 2002-02-23 at 07:09, Peter Dimov wrote:
> > > I think that this discussion would be more productive if it's based on
> > > concrete examples. Saying "Boost doesn't install", "Boost users", "Boost
> > > this" and "Boost that" borders on abstraction.
> > >
> > > Which concrete libraries are your users having trouble with? Complain to
> the
> > > library maintainer directly. Avoid the "Why Jam" issue, concentrate on
> "Why
> > > doesn't library X build."
> >
> > Peter, this has already been covered at length: my users expect a sane
> > install target; Boost doesn't have one. It's that simple. It's not about
> > building libraries, it's about installing them (and their headers).
> 
> I don't think that "this has already been covered at length." Everyone
> participating in the thread obviously has different ideas about what has
> been covered. :-)
> 
> If it's not about building, why is the subject "Why Jam"? Jam builds.

As I see it, these are the issues:

        * There are scalability problems with a configuration header of
          the ilk Boost uses. These scalability problems are why
          autoconf exists. It is apparent at this point, though, that
          the Boost maintainers do not feel that they're hitting the
          ceiling with the configuration header approach. So there is
          not motivation to move to autoconf or a similar tool on this
          ground. IMO, it's entirely appropriate to dismiss arguments to
          move to autoconf on this basis.
        * Jam is Weird. On POSIX systems, the autotools have come to
          provide the closest thing to a standard build environment as I
          think anyone might reasonably hope for. This has made us soft.
          Hostility toward an "alien" build setup--which makes people
          learn New Ways of doing things when the Old Ways seemed to
          work Just Fine--while inappropriate, is probably to be
          expected. But these arguments can also be dismissed, as they
          are founded upon ignorance.
        * If my open source project depends on Boost, how am I supposed
          to tell my users to install Boost? Some persons here seem to
          think it is appropriate to ask every open source project out
          there to roll its own Boost distribution. I strongly disagree
          with that. The Right Thing is to solve the problem once, at
          the source. The autotools are one way of doing that. I have
          not seen anyone produce an alternative strategy for solving
          this problem that is workable in the near term.

-- 
Braden McDaniel                           e-mail: <braden_at_[hidden]>
<http://endoframe.com>                    Jabber: <braden_at_[hidden]>