Re: libc strategy
On Mon, Jul 02, 2001 at 07:45:05PM -0500, GT wrote:
> Quoting Nathan Myers <ncm@nospam.cantrip.org>:
>
> > Such a library already exists in BSD, and need only be packaged for
> > Debian. In effect, (c) is just an implementation detail of (b) that
> > allows much of the porting effort to be shared among all the ported
> > packages. A person willing to put more work into a port might bypass
> > the compat library.
> >
> > Actually, the BSD compat library even provides a degree of binary
> > compatibility, which we don't need. Much of it could be discarded.
>
> If Nathan's right, I don't see why we couldn't start
> producing transitional packages Real Soon Now(tm).
"Now(tm)" would be a good time. Anybody with a Summer Vacation can
go ahead and start packaging things Right Now. There's no need for
more talk; it's clear what has to happen, and either somebody just
does it, or the project will have to be considered vaporware again.
To get started: if you don't have NetBSD, install a minimal system.
It's easy.
Start Debianizing core BSD kernel-related packages. Port dpkg, apt,
and the package build tools. Start replacing raw binaries by unpacking
your Debianized versions. Phase 1 is complete when all the binaries
in the minimal system were unpacked from .deb files.
For Phase 2, port the Debain installer. For phase 3, start building
non-kernel-related Debian packages for the system, patching as you go.
(Phase 3 is never done.)
Don't worry about duplicating somebody else's work; the best of both
will end up used, and at worst you'll learn a lot. There's no need to
decide whether to use BSD libc or GNU libc; if you want to try porting
GNU libc, just go ahead. If you do a good job your work will be used,
and adopted into the upstream package; the GNU libc project welcomes
ports. Debian GNU/BSD can use as many libc's as there are. (Cygnus
newlib might be a good idea too.)
Nathan Myers
ncm at cantrip dot org
Reply to: