[boost] Re: Re: Proposal: Identifying libraries

From: Edward Diener (eddielee_at_[hidden])
Date: 2002-12-18 00:53:14

"Beman Dawes" <bdawes_at_[hidden]> wrote in message
news:4.3.2.7.2.20021217132539.01ee10b0_at_mailhost.esva.net...
> At 08:29 PM 12/16/2002, Edward Diener wrote:
>
>  >"Beman Dawes" <bdawes_at_[hidden]> wrote in message
>  >news:4.3.2.7.2.20021216085952.01b42678_at_mailhost.esva.net...
>  >> We've got two cases where it would be very useful to be able to easily
>  >> extract the library name from a header file:
>  >>
>  >> * Generating a dependency analysis based on library names rather than
>  >than
>  >> header names.
>  >
>  >Thanks. I had suggested such a dependency tree based on libraries rather
>  >than on headers be available for those who want to distribute Boost
> without
>  >having to distribute the entire header file directory. Of course each
>  >library should have a list of its headers in such a dependency tree. I
am
>  >glad you have picked up on the practicality of that idea.
>
> There have been a number of ideas floated, such as a dependency tree based
> on libraries, that I wanted to tackle in C++ rather than one of the
> traditional scripting languages. Hence the detour to do the filesystem
> library.
>
> Having a relatively easy way to identify libraries will further ease the
> programming.
>
>  > As much as I like
>  >Boost I have been reticent to use portions of it in distributable
library
>  >code since all the headers must be included in the distribution, unless
I
>  >worked out all the header dependencies for all areas of Boost which I
> use,
>  >which is too much of a PITA even to contemplate. With dependencies based
> on
>  >libraries, and with all of the header files in a particular library
>  >specified in such a dependency tree, distributing just portions of Boost
>  >will be much easier.
>
> Perhaps it might be possible to use the dependency information to drive a
> download process that only pulled in what was asked for, plus
> dependencies.  I think cygwin does that now in their setup wizard.

That certainly sounds like a good possibility also. I admit personally that
I am on a high-speed connection so downloading all of Boost is not a big
deal for me. But being able to both download and distribute only the
libraries which one needs from Boost could be an important issue. For myself
it is an important issue to distribute only what I use, because I don't want
to burden the end user with extraneous files even if the extra size of all
of Boost's headers is not significant. I can't speak for other programmers
who use parts of Boost in their implementations and are distributing just
those parts for their own libraries. It is, of course, possible to
distribute code which uses parts of Boost without having any need to
distribute Boost header files or libraries. My own implementation needed
some parts of Boost in order to support advanced C++ functionality of which
the end-user programmer could take advantage and so I distributed the full
set of header files rather than the parts I wanted to expose (
boost::function and boost::bind ). I imagine that there are other
programmers who might need parts of Boost for their own 3rd party library
and must distribute those parts for their end-user programmers to take
advantage of functionality in their library.