On Wed, Apr 10, 2002 at 02:07:27PM +1200, Volker Kuhlmann wrote:
> > That's one of the problems with making your package management system
> > file based, instead of package based.
> The problem with making the package management system package based, of
> course is that everybod calls their packages different names, although
> they contain the same software.
No problem. See 'Provides:' functionality in dpkg. Also take a look at
the idea of "fake" or "virtual" packages.
> You can't win either way. Actually, the file-based approach is more
> reliable in case of libraries, which contain a version number in the
> file name, and package dependencies can be set up as >=version.
I disagree.
The filename of a library is not a unique way to identify that file -
creating a package that depends on libcomplex.so.4 only guarantees that,
with satisfied dependencies, you have some file named libcomplex.so.4.
Using a alternative package format, based on package dependencies, you
build the original package to depend on foo-mathslibs, knowing that
foo-mathslibs includes the correct libcomplex.so.4 - i.e. the one it was
linked against. There is still a chance that someone can create a
package with the same name which includes the wrong libcomplex.so.4, or
even lacks that file at all - but the chance of that happening is lower
than of a conflict at the file naming level.
> apt does not solve this problem either, Debian just has a
> rigorous(true?) system of package naming. Any independent (as opposed
> to "based on") apt-distros (are there any?) presumably simply copies
> the naming scheme, thus sidestepping the problem, not solving it. The
> naming scheme is the real advantage of apt over rpm, not its technology
> (which is IMHO worse). The same could be achieved with rpm, and there's
> no guarantee it'll always work with apt. It'll always work with Debian,
> yeah fancy that, it'll also *ALWAYS* work with Red Hat, and SuSE, and
> etc ...
Firstly, I'm not sure why you're talking about apt - do you mean dpkg
and '.deb' packages? apt is just a front end to the dpkg system.
The naming scheme is one part of why the dpkg system works better, but
there are other reasons why it is superior to RPM, too. Also, the naming
scheme for Debian packages is well documented, so anybody going to the
effort of building a dpkg should be following this scheme.
The main advantage of RPM is its ubiquity, not its technological
superiority to other package management systems, in my opinion. I am
curious as to why you think RPM is technically better than dpkg.
dpkg is not perfect, and that is no secret - read the debial-devel
mailing lists - but it is better than RPM. ;-)
--
Matthew Gregan |/
/| [EMAIL PROTECTED]
