Re: apt management (was kitchen)

[Heather <star@betelgeuse.starshine.org>]

>  Problems with unstable are mostly in dependencies and pre/postinst scripts,
> and other minor packaging errors. 

...which can really make a bad hair day out of your dpkg setup.

> What I do is run woody, but with the unstable repositories in my 
> sources.list, and APT::Default-Release "testing";  in my apt.conf. 
> (see apt_preferences(5), etc.)  

what I have is different sources lists, because I like to see what versions
are in testing or stable or even progeny (on behalf of others) and then go
back to stuff that follows Potato for normal use.

> This way,  apt-get install package   gets the package from woody,
> unless it only exists in unstable.  apt-get install package/unstable  gets
> the unstable version.  apt-get -t unstable  lets apt upgrade the
> dependencies to their unstable version if necessary.
 
Hmmm, it'd be interesting to have that only fetch source debs from anything
"later" than what you're tracking.  Then it'd actually act more like I do.

>  I've found that running an unstable libc is not actually as bad as it
> sounds.  It's a really important package, so Ben makes sure he doesn't break
> it :) 

Heh, it's all those *other* thingies making assumptions about it, that I have
the problem with, but staying off the new libc keeps them from trying to sneak
into my kit.

> Of course, I wouldn't be running unstable packages if I didn't think
> I could hack my way out of any problems, so it's not for novices, but
> having a few packages from unstable is not a big deal, 

If it doesn't care what libc it needs, I'd probably try it (if it relates
to what I do with the laptop)

> Running the unstable versions of packages you are actively interested in is
> a good way to contribute to debian, 

Yes, if you're willing to swat bugs, you should definitely play with unstable;
the more the merrier.

> and takes a lot less effort than hoping
> that lots of people will test everything by running unstable across the board.
 
> (Notice that I managed to argue that my position is the Right Thing to do.
> Err, I didn't quite mean to do that, but whatever. :)

Oh well.  Different things are the Right Thing To Do for different people.  
Requirements analysis (what's needed, what are you willing to risk) and
Desirements analysis (what would be perfect, if the world were ideal) are both
worth looking at before flat-out making recommendations for other folks, tho.

We now return to your regularly scheduled laptops.

* Heather Stern * star@ many places...