Trackin' stable vs. releases

[Rick Moen]

Quoting Alan DuBoff (aland@softorchestra.com):

> There was an interesting topic that Rick brought up at Cafe Borrones last 
> night, that being what do releases really matter if you're trackin' stable?

Well, to put it another way, I was trying to think of how to most
effectively explain the Debian way of doing things to people (a category
that on a good day might also encompass reporters).  I said one tack
might be to say, in essence, there are exactly three Debian branches:

stable
testing
unstable

They incrementally improve daily, and you can get point releases as
snapshots, but those _are_ Debian's "releases" in (almost) any sense
that matters.

I.e., I said picture a guy who installed 1.3/hamm when it was the
current stable branch.  His sources.list says stable, so at intervals he
resyncs.  Packages gradually improve.  The system slowly changes.  But,
each and every day, he's running the Debian-stable release, current as
of his last resync.  He hears news about "releases", but they aren't 
relevant to his system.  From his perspective, it doesn't matter if
Debian ever "releases" again.  He'll just keep on running Debian-stable,
whatever Debian-stable is at a given time.

One day, our guy hears that there's a controversy in Linux Weekly News
over whether Debian will keep issuing security updates to "potato".  He
checks the Debian FAQ as a reminder, and finds out that until June 2002, 
"potato/2.2" was Debian-stable, but then got shuffled off to
debian-archive as obsolete.

He scratches his head and says "Wait, I don't get it.  There were timely
security updates to Debian-stable before June 2002, and there were
timely updates to Debian-stable after June 2002.  What exactly is the
problem?"

The problem, of course, is that the LWN guy (if slink experience is any
guide) doesn't get that Debian-stable _is_ the release in all senses
that matter, and people who insist on continuing to track a release
after it's discontinued just aren't clear on the concept, and is saying 
"I want my system to be decrepit and unmaintained".

> This is certainly an interesting topic....

Well, I wouldn't count on it.  We may have just bored everyone to tears.

> ...but I find myself at some point in the cycle, moving to unstable
> for various reasons to do with the newer software that I prefer to be
> using is on unstable.

As I said at the meeting, this is a different subject completely.  It has
nothing to do with the concept of "releasing" Debian and the role of
functional track names (stable, testing, unstable), and is equally true
with or without formal releases.  In fact, it would remain true if the
Toy Story names and version numbers were abolished entirely, and
everything were just referred to as stable, testing, unstable -- drawing
from packages in the pools directory that change over time.

> As I have pointed out, for me there is usually software that drives my 
> decision to move to unstable, and in the case of woody, I'll stick on stable 
> for quite a while after woody is released. 

But the question is, will you stay on _woody_?  Presumably, your
sources.list currently says "stable", and you occasionally pull packages
from unstable.  When woody goes stable, the symlink gets remapped on the
mirror sites, and your system will get auto-upgraded to woody (because
it'll be the new stable).  But the only reason why you'll be on woody
will _be_ that it's the stable branch.

If you intend to use (in sources.list) specific branch names like potato
or woody, or do so now, I'm curious about why.

I'm sure there are sometimes compelling reasons.  Say, someone decides
that he really likes the basic architecture of slink, and selectively
builds and upgrades packages manually where he needs to.

> However, it was software like XFree86 4.x, Mozilla, and even <gasp>
> Emacs 21.1 that make being on unstable satisfying for me at some point
> in the cycle.

One wonders why you aren't on testing?

-- 
Cheers,         (Regarding "In God we trust":)  "Don't ask me how you set the
Rick Moen       trust-level of a god."  "At a PGP signing party?"
rick@linuxmafia.com -- Per Leijonhufvud & Peter de Silva, in ASR