Re: Activating t-p-u by default (was: Re: For those who care about their packages in Debian)
2010/8/26 Carsten Hey <firstname.lastname@example.org>:
> * David Kalnischkies [2010-08-26 17:43 +0200]:
>> Long story short:
>> If you want to get updates from an archive only if you pushed a version
>> previously from it: 100 => pin > 500.
> Wouldn't adding a new field to Release files similar to 'Not-Automatic'
> but pin to 101 instead of 1 if this new field is set to yes an option
> for apt/Squeeze+1? This has been reported in #186767.
Well, yes and no.
Technical more or less no problem, but…
As far as i understand t-p-u i don't understand why the default should
be < 500. If i am adding it (or let a piece of software add it for me)
i guess i want these packages on my system: proposed-updates at least
suggests for me, that they will be soon in testing anyway, just,
that they are tested now before they enter "real" testing in an overlay
archive. If i don't want to participate in this testing,
i should remove the archive from my sources.
If i just want to grab some versions from it, t-p-u doesn't get much testing
in general as you must care for this specific package enough to get a new
version for it: So new iceweasels are maybe tested a bit, but packages like
tzdata will not get any testing through t-p-u…
The problem with this is also, which is why i don't think it would be
suitable for backports, is that these archives mixing minor only-bugfix
releases and new groundbreaking upstream releases…
E.g.: I maybe want to get bugfix releases for iceweasel through backports
automatically, but what i don't want is an automatic 3.6 -> 4.0 upgrade,
but such pinning i need to define by hand anyway.
Regarding the bug: What do you do if two non-debian archives provide such
a package -- and, do they "fight" against each other now that they can by
changing their DefaultPriority? The cleaner way for the user would it be to
declare: I don't want to get this package from this archive ever and
i care only for these packages from this archive, ignore the rest.
You can say that with -1 and Co, but it is a bit harder than needed…
So, again in short:
I don't see a reason backports shouldn't be pinned to 1 by default and
t-p-u by default not pinned at all to get the default 500 pin…