Re: suggest renaming and extending the -CURRENT and -STABLE lines

On Thu, 11 Oct 2007 01:58:51 +0200 Ivan Voras wrote:

> Unfortunately (for this developer-centric practice), the trend in large
> and/or important production environments is - as seen in Linux (and
> Solaris) - to severely limit major OS upgrades. 

I don't really see what's so unfortunate about that. Most admins
(including myself) go by the rule "If it ain't broken, don't fix it!"

Production systems, especially in companies are there to do something, and
that is not trying out the latest new gadgets out there. In my company,
out database server is *very* important and although downtime would not
cost us any money off hand, we wouldn't be able to serve our customers
anymore, since the information stored in there actually is the service we
provide.

I am very careful about trying out everything new out there, because
every update bears the risk of downtime. And both -STABLE, but especially
-CURRENT have risks of that sort.

> Of course the existing possibility to do is excellent, but more and more
> end-users, especially big ones, are going with big Linux distributions
> that basically stay frozen (except for security upgrades) for years. And
> this idea gets a +1 from me - productions releases that are expected to
> run for years should be able to "just work" without upgrading to the
> "kernel of the week". To do this, a stable anchor-point is required and
> that's what -RELEASEes should be for. I think the fact that not all our
> -RELEASES are created equal (some are "extended support" releases)
> should be more advertised and explained.

I'm not even sure we should board that train completely. Basicly,
RELENG_x_y is just that, but is could also be a good idea to have and
alternative. Something that is stable but can still be updated. Usually,
an update from one release to the next is relatively painless (major
releases excluded). I kept a machine running on RELENG_4 from 4.0-RELEASE
until now without ever having to change or reinstall the packages on it.
So it is possible. The next step would be to have something that updates
like RELENG_x but only uses tested and stable code like RELENG_x_y without
confining the user to a single release.

Regards
Chris