Re: What is the use case for Policy §7.6.2 ?
On Fri, Nov 9, 2012 at 10:43 AM, Josselin Mouette <joss@debian.org> wrote:
> It looks to me that we should strictly favor the transitional package
> approach instead.
Amen (with any of my hats on).
> Shouldn’t we entirely forbid the
> Provides/Conflicts/Replaces combination way of handling upgrades, except
> for virtual packages?
With my non-D{D,M} hat on just reading the debian-policy I don't see where
it is talking about upgrades here. The section is talking about replacing
packages with other packages - not that this should be done automatic or
should even be a suggestion from a package manager.
With my APT hat on I don't see how "upgrades" should be implemented in
that context which do not replace exim4 with postfix just because both are
mail-transport-agents. You could probably work around it by saying that the
replacee can't replace the replacer (and v.v.) but then you have the problem
of deciding which version of the replacee should be considered (installed,
candidate, the one in NEW, …), not to mention the potential social problems
coming from a iceweasel provides/conflicts/replaces chromium upgrades. ;)
More seriously it probably will not work anyway as somewhere is a versioned
dependency in the mix, so you are lost anyway just that many people doesn't
know it (and don't say that is not true, I seen people trying to replace
libaries that way …)
The syntax is fine for what policy defines, it just doesn't provide what
many people read into it.
Best regards
David Kalnischkies
Reply to: