Doesn't a new version replace the previous one?

[Jeff Spaleta]

On 6/23/05, seth vidal <skvidal at phy.duke.edu> wrote:
> Mostly due to the lack of a program being run to prune them out. I've
> got the program and we run it from time to time - but I think there's
> some dispute over whether it should be run everytime or not.
> 
> your thoughts?

for the record.... for the one person left on the planet who hasn't
heard my opinion yet.

I really really think we should leave an older version in a publicly
available tree long enough to make sure a drop dead stupid regression
on a package update doesn't burn users.  Nobody is perfect and there
will be forehead slapping mistakes that sneak through and cause
torment for users.  Its much easier for those of us who try to help
users identify problems and find workarounds if we can instruct them
to do a manual dowgrade when this sort of things happens.  Anybody who
remembers trying to use ximian on top of rhl 7.x will remember the
joys of the fully rolling release tree xim had that contained no older
packages.    I'd really like to avoid any user being sort of stuck out
in the cold if an update comes down with a problem.

Hold an older version for more than a  day... maybe a week tops. If
the maintainers havent noticed in a week that there is a big
problem... we have other things to worry about.

-jef