Re: Let's CENSOR it! (was: Uploaded anarchism 7.5-1 (source all) to master)
On Thu, 25 Mar 1999, Josip Rodin wrote:
> On Thu, Mar 25, 1999 at 12:29:56PM +0100, Richard Braakman wrote:
> > > Indeed, I understood meaning of Replaces: as 'replaces files from
> > > named package' not neccessarily as 'replaces the whole named package'.
> >
> > The Replaces: header carries both meanings. It's the second if the package
> > also Conflicts: with the named package.
>
> I didn't say otherwise. Still, two packages may have Conflicts: and
> Replaces: on the same package, and AFAIK there is no AI in dselect
> or apt that could determine which one to automatically install, and
> we shouldn't expect that from them.
Exactly, "AI" is the perfect term for this :-)
I fully agree that we should not expect dselect to have AI from the
current and existing dpkg fields.
> For example, all MTAs conflict and replace other MTAs (through
> mail-transport-agent), but they can't be treated in the same manner
> as these renamed packages.
>
> That is why an additional, decisive, lucid, field is needed, that
> would enable renaming and obsoleting packages automatically.
A new field is needed to do in an *elegant* way (and of course a new field
with the exact desired meaning is much better design that doing AI).
> Or some other method?
Until a new field exist we can use dummy packages, which are functionaly
equivalent, and do not break any policy (no need to fiddle with dpkg
internal database by maniuplating /var/lib/dpkg/info directly, for
example).
After an upgrade, the user usually ends up having obsolete libraries
which are no longer needed. Well, an empty package does even less harm
than that, because at least it does not take any disk space. I don't see,
therefore, why a dummy package is so "evil" and "uglier than hell".
Thanks.
--
"fbd86064ca7511149a785838ae2cdfc0" (a truly random sig)
Reply to: