Availability of "unstable/interim" CDs (Was: Danger Will Robinson ! Danger!)
From: Jacob Kuntz [mailto:jake@megabite.net]
>
> our biggest handicap is that we're always a year behind everyone
> else. being a year behind is suicide in any industry. being a year behind
> in an industry that moves as fast as open source software, is idiocy.
Agreed. With hardware changing as rapidly as it does, you end up with a
situation where relatively low-end current hardware is unsupported (or at
least badly supported) by the released version of debian. Graphics hardware
and X is the worst example, but things like USB/APM/PnP are similar.
> our stable release is using 2.0.36. most people are afraid of
> our 'unstable' tree.
Not necessarily afraid. However, the only version I can get on CD is slink.
Yes, I can upgrade over the net, but I pay for connect time, and the cost of
downloading a full potato upgrade is not trivial. Even less so when I have
to repeat the process regularly. Add that to the fact that I have to do the
downloads again to upgrade my other PC....
What would be REALLY valuable would be:
1. Easily available, regular "unstable" snapshots on CD. They don't have to
be guaranteed reliable, they don't have to be supported, but for many users
they are pretty much a necessity to run debian on recent hardware.
2. A good way of doing a one-off download, burnable onto CD, of a tailored
"upgrade" subset which covers the important packages. The user should be
able to define "important" (what's currently installed, plus manual
selection is a good start). Automatic dependency following is essential. The
resulting set of files should be sharable via network, zip drive (where
small enough), or burnable onto CD. Being able to incrementally refresh such
a subset would be an added bonus.
Of course, if either or both of these things are already available, then
what is needed is to give them a higher profile...
> you've seen all the threads about people trying to upgrade from
> slink to potato and having all sorts of problems. why do they do it?
> because slink is so far behind that it isn't usefull anymore.
Look at it another way. Slink is great. It's highly reliable. One of the
good things about Debian is its QA - I have no problem with telling people
that if they want a rock-solid Linux, then Debian is the place to go.
But for people who are willing to accept a *slightly* lower level of
reliability (and there must be lots of them, or how do you account for the
popularity of Red Hat? :-), the Debian unstable distribution is a very
attractive choice. For example, I don't have any mission-critical services
on Debian, but I am a "latest version" junkie. So, for me, unstable is the
"right" choice. But short of a CD image download, or a slink+net-upgrade
approach, I can't get it.
> IMHO, leaving out 2.4 is a bad idea.
I disagree. The approach taken by slink was sensible. Have 2.0 as the base,
because it was QA'd to the high standards required by Debian, but include
the latest 2.2 source package for people willing to upgrade. Adding a bit
more support, in the form of including the equivalent of the "Using Kernel
2.2 with Slink" webpage on the CD, and including (in a separate directory)
debs for the relevant "unstable" versions of packages which need upgrading,
would be enough.
> this thread brings up an interesting topic: how can we keep up?
Agreed, Debian is by FAR my favourite distribution, but the slow release
cycle is a real problem.
> i think i have the beginning of a good idea. please flame/comment
> as you see fit.
>
> make a release every 3 months with an official cd image,
> fanfair on the website, the whole shebang. only include enough
> on the cd to do a basic install. only consider 'release critical'
> bugs release critical if they're against required base pacakges.
> the rest of the distribution would remain on the archive sites.
>
> with this pattern, we produce four releases per year. three
> interim releases (2.3, 2.4, 2.5) and one major release (3.0).
> in order to figure out what packages to include on the interim
> release, we probably should get statistics on what most people
> use. perhaps analize logs from the archive sites, and encourage
> more people to use popularity-contest.deb.
>
> what do you folks think?
I disagree. Adding to the QA workload is likely to be fatal. And your
interim releases still require QA-ing. Instead, I suggest making "official"
releases of unstable. The "stable" release remains the baseline supported
version for people with mission-critical requirements, or for whom stability
is more important than the latest versions. However, the unstable snapshot
becomes an alternative for people who want latest versions, but who can
accept the risks. Add a minimal amount of "packaging" - a release note, plus
a short period of critical bug squashing (no more than a week or two, with
unfixed critical bugs simply being documented in the release note) and you
have a useful resource...
To make the point, such a release could be called an "Official Debian
Unstable Snapshot" or something similar - to give the flavour of official
sanction without any guarantees :-)
Whether this makes any real sense, I don't know, as it's not clear to me how
the existing "official" CDs get published though the existing resale
channels (CheapBytes, etc).
Just my thoughts...
Paul Moore.
Reply to: