Re: Upgrading Squeeze to SID
On Mon, 29 Aug 2011 12:26:30 +0100
Lisi <firstname.lastname@example.org> wrote:
> On Monday 29 August 2011 12:05:47 SZERVÁC Attila wrote:
> > R U a Debian Developer?
> > If not, DON'T upgrade to *UNSTABLE* ('sid') distribution. 'sid' is
> > the *UNSTABLE*, *very buggy*, *INCONSISTENT* distribution for
> > *Debian developers only* - if U want new packages, use Debian
> > Backports or *testing* (wheezy).
> That is a little harsh. Many people run Sid who are not DDs. Only
> run Sid if you are adept at sorting out problems and can cope if a
> package (perhaps one you use daily) is broken for a while: days, or
> perhaps a week, not months or years.
I would also say that was an exaggeration. If you have only one
computer, and rely on it totally, (a so-called 'production' system) then
I agree it is risky to run sid. Even if you know how to fix a breakage,
you may miss a deadline or lose important work. But if you have
multiple computers, or virtual computers, or you dual-boot, and you
have sensible data backup measures in place, then sid is relatively
You do have to learn how to fix things, but that's always the case when
running Linux. Even Ubuntu needs a bit of a kicking now and then, and
testing is a long way short of Ubuntu in terms of spoon-feeding.
I have used sid for the large majority of work that does not use
proprietary hardware for over five years, during which time I have
needed to reinstall twice, due to insufficient knowledge to fix the
problems in a reasonable time. Since it is only a workstation, and my
data lives on a server (running stable), reinstallation was relatively
painless, and was fundamentally similar to what the OP wants to do i.e.
throw on a stable netinstall, change sources and dist-upgrade then
restore the previous packages.
There are a few sensible precautions in addition to data backup: keep a
fairly up-to-date package list, a full backup of /etc and a listing of
disc partitions, filesystems etc. But even if sid does break beyond
your ability to repair, you can almost certainly recover this
information from the system. It's not like a dead Windows installation,
which is normally unrecoverable unless it is a simple boot issue.
There is also a certain inconsistency in claiming that testing is only
ten days behind a 'very buggy' and 'inconsistent' distribution. Serious
problems don't get fixed that quickly, or at least not every time. And
testing will spend about the last half of its life frozen, getting no
further major upgrades and falling further and further behind sid until
the next release, at which point life in testing also becomes very
interesting for a while.