[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: question for sources.list

On Sun, Sep 04, 2005 at 09:24:21AM -0400, Hendrik Boom wrote:
> On Sat, Sep 03, 2005 at 02:32:11PM -0600, Paul E Condon wrote:
> > On Sat, Sep 03, 2005 at 11:46:19AM -0400, Hendrik Boom wrote:
> > > On Fri, Sep 02, 2005 at 10:40:40PM -0600, Paul E Condon wrote:
> > > > 
> > > > So, until some time in the far future, people should not say to
> > > > newbies that release code names and release status names ('stable',
> > > > 'testing', etc.) are interchangeable. They are not. Existing support
> > > > for release code names is, in fact, quite restricted by comparison.
> It looks as if the advice I gave below may only be valid for
> sources.list.  I apologise if I was misleading, but see further discussion.
> - hendrik
> > > 
> > > People that want stability had better use the release code names, otherwise
> > > they will suddenly be dragged into the future every few years and
> > > have little control over it.
> > > 
> > > People who want frequent releases had better specify the release
> > > status name (testing) or every few years their system will gradually
> > > start going out of date.
> > > 
> > > People who break toys can use either.  Aren't "sid" and "unstable"
> > > really interchangable?
> > > 
> > 
> > Yes, but that is a vacuous test. Sid/unstable _always_ contains the
> > packages with the highest version number. It _never_ needs to be
> > specified as the preferred release, if it is the release that you wish
> > to follow. On the other hand, if you wish to follow mainly testing,
> > and have sid/unstable in your sources.list so that you can pick up a
> > few packages before they reach testing, using 'etch' in either
> > apt.conf or apt/preferences does _not_ prevent the automatic upgrade
> > of _all_ packages from testing to unstable. IMHO, very few newbies
> > really want to track unstable, though some may want to grab one or two
> > packages from it.
> > 
> > When you do grab a package from sid/unstable, you must use 'unstable'
> > on the command line to apt-get (you might think you can use the version,
> > but sid/unstable does not have a version number)
> > 
> > In short, even for sid/unstable, the release code name and the 'Archive name'
> > are _not_ equivalent.
> > 
> > It is easy to run your own tests of this. If it matters to you, you should 
> > run your own tests. The wording of the documentation is difficult to 
> > understand, but not incorrect. If doesn't matter to you, and/or you have
> > not run some careful tests of your understanding, you should refrain from
> > giving advice.
> Thanks for the correction/clarification.  I thought I understood,
> but now I know I don't.
> Is it that "sid" and "unstable" mean the same on /etc/source.list,
> because the two names are permanently paired off
> (unlike "woody"="stable", which became false recently),
> but they have different meaning or validity in apt.conf
> and apt/preferences?

In http.debian.org, softlinks are used to pair stable with sarge,
testing with etch, and unstable with sid. The last time I looked (at
least a few months ago) unstable was the real directory and sid was
the softlink. Today, it is the other way around: Sid is real directory
and unstable is a softlink. It appears that some day soon all the
nasty complication of which I have been warning will be gone. There is
still a complication that may reach out and bite someone: the header
of the file named Release, for sarge contains two lines,

> Suite: stable   
> Codename: sarge

These two lines are consistent now, but will be inconsistent when etch
becomes stable. Maybe this will be gone soon, too.

Paul E Condon           

Reply to: