[GTALUG] Virtualized OSes... or perhaps pointing out the obvious problems in NixOS

D. Hugh Redelmeier hugh at mimosa.com
Thu Apr 23 19:30:07 UTC 2015


| From: Lennart Sorensen <lsorense at csclub.uwaterloo.ca>

| Certainly the problem it is trying to solve is not one that has ever
| really been an issue for me (the few times I have had a reason to use
| another version of something, I have installed it in a small chroot
| running an appropriate version of the distribution to run that version).

Me too, I think.

Sometimes "systems" get stuck on old versions of Linux (let's forget
other OSes for now).  Why do they get stuck?  I think that they are
working and changing to a new distro release has a significant chance of
destabalizing them.  Maybe the process to actually configure the old
system is lost.

But you don't want everything in your organization stuck at that old
version.

So you end up with a diverse set of versions.  They cannot normally
live on one box.  The cure is virtualization or some other trick.

This all seems wrong, but all too understandable.  (I have a couple of
boxes still running RHL 8.0 from 2002.)

So what are the correct solutions?

- RHEL has something like 10 year stability for each release.  Ubuntu
  has LTS.  This allows you to safely ignore the problem for a decade.
  By which time it is enormous!  But of course they don't force you
  to ignore the problem.

- keeping up with the rat-race of rolling updates is noble but probably
  wasteful.  It is the opposite approach to LTS.

- virtualization allows you to put off the problem for a long long
  time.  But it doesn't address vulnerabilities in no-longer-supported
  distros.

- For me, running a virtual copy of an ancient system let me avoid
  buying a new license when the old machine wasn't worth running any
  longer.

- NixOS would seem to address this.  Is that its justification?

What do real businesses do?  What should they do?


More information about the talk mailing list