Re: RPMs, Specs and some noob questions

[bob proulx com (Bob Proulx)]

Tony Earnshaw wrote:
> Bob Proulx wrote, on 12. mar 2007 05:29:
> > > RPMs seem like the best way to keep version control, especially because 
> > > it's such a widespread protocol.
> >
> >tar.gz files are even more widespread and popular.  :-)
> 
> I used to think like that before I got into serious sysadminning (even 
> you don't really think that, you're just provoking), but installing 
> often highly dependent stuff and configuring it on multiple systems with 
> a myriad of custom-written configuration scripts calls for versioning 
> and I haven't found anything better than rpm yet (old SCO, Novell 
> UnixWare, SGI and Solaris admin).

Lest I am completely misunderstood let me clarify that I am strongly
an advocate of packaging software into rpms.  I actually think are are
pretty much in agreement there.

The part that I don't like is rpms that simply transport files and
then make a mess of things in the '%post' script.  I think a bad rpm
is worse than no rpm.  Perhaps that is our point of disagreement.  If
an rpm is nothing but a facade to fool the casual observer then I
think that is a bad thing.

I expect rpm installed files to 'rpm -qf /path/to/file' report the
package that installed them (within reason and good judgement).  I
expect 'rpm -ql' to list the files in the package.

I expect 'rpm -V' to be able to verify that the package is still
installed cleanly.

I expect 'rpm -Uvh' to proper upgrade to a newer version without
problems.

I expect 'rpm -e' to clean up properly.

I expect that if I install the same rpm next week that it will install
exactly the same way that it installs today.  Repeatability and
reliability matter.

Bob