Re: Use with Ximian RPMs?

On Tue, Jul 16, 2002 at 02:16:04PM -0400, Peter Bowen wrote:
 
> No, I would consider it to be a bug that a sort function is not
> transitive.  rpmvercmp can get it self into a situation where pkgA >
> pkgB _and_ pkgB > pkgA.  This is a bug, IMNSHO.

True, and I agree with you - however playing devil's advocate:

Is this a format bug or a policy bug. At the end of the day RPM is a
fairly well spread format, and actually in some ways that sucks.  It has
good points - cryptographic and hash verification of packages, md5 db of
files, etc. As a sysadmin and sometimes packager this is good. 

However naming and numbering is policy.  I tend to draw upon a vendor
tree, eg RH, and merge my stuff in as rpms based around a policy based
thing.  However, if I have multiple upstream providers how can I
determine preference.

As far as I see current's role it's to be a provider of a vender (or
managed set of vendor's) tree and updates.  It's not psychic.

Yes, there needs to be either more standardisation or more awareness of
packaging and versioning, and yes it'd be nice if people packaged better
RPM's.  I tend to grab a src.rpm for an external package and make it
work for me, which may be rpm -ba, or may be a patch or may involve me
banging my head against a brick wall (plptools is currently annoying
me!).

Anyway I'm wibbling and probably had a point but it's lost

Paul


> 
> Thanks.
> Peter
> 
> 
> _______________________________________________
> Current-server mailing list
> [email protected]
> http://lists.dulug.duke.edu/mailman/listinfo/current-server