On Sun, 2002-02-24 at 15:39, Peter Eisentraut wrote:
> Karl DeBisschop writes:
>
> > It's one thing to as people to run a test. It's another thing still to
> > ask them to run that test when the only readily available test method
> > bypasses whatever administrative packaing policy is in place on those
> > machines.
>
> It seems unlikely that many would want to install a development snapshot
> as their main installation.
Depends what you mean by main installation. I have a few Redhat servers
whose only reason for existing is to test software. So would I want it
to be my main installation for any production-related purpose? No. But
would I want to track PostgreSQL development on that server? Sure, why
not? At least why not if I can use RPM to cleanly remove the software.
Our company policy is that wherever possible, the native package
management tools should be used to install software on a machine,
whether it's a development box or a production server. RPMs are the
native packakging format for Redhat, so we hope to install software
using RPM. I might add, particularly if it's not production software --
since RPM then helps assure that everything that should be removed can
be removed.
Now it is true that I can just install it to part of my home directory,
but that's less than fun over NFS.
And the fact of the matter is, if the goal is to expand testing, which
was what I was trying to comment on, you probably get alot of mileage in
expanding that testing by making packages easy, RPMs, BSD packages,
solaris packages , whatever. It's just a fact that there are people out
there who have little interest in the guts of compilation, but still
have the ability to install and burn in an alpha or beta release.
Again, just my 2 cents.
--
Karl DeBisschop <kdebisschop@alert.infoplease.com>
The Learning Network / Reference
www.learningnetwork.com / www.infoplease.com
