Re: How bad is a wrong recommends? (was: Bug#273734: education-common: con't fulfill the Recommends on !i386)
On Sun, Oct 03, 2004 at 07:03:50PM +0200, Andreas Barth wrote:
> * Adrian Bunk (adrian.bunk@stusta.de) [041003 17:10]:
> > On Sun, Oct 03, 2004 at 12:44:21PM +0100, Martin Michlmayr wrote:
> > > You certainly have a good point here. I'm not suggesting to remove
> > > Recommends; I think the concept of Recommends is good. However, there
> > > is also a difference between Depends and Recommends. I think what I'd
> > > like to see is:
> > >
> > > Depends -> grave bug
> > > Recommends -> normal (or important) bug
> > > Suggests -> minor
> > >
> > > Recommends is stronger than Suggests but it doesn't completely break
> > > the package so imho it shouldn't be RC.
>
> > It depends on how strong you expect a Recommends to be. My impression
> > was, it's a "install the recommended package unless you really know what
> > you are doing".
> >
> > And it's currently supported that in order to aid users a package
> > management tool might handle recommends like dependencies. If this
> > should continue to be supported, they have to be treated the same way.
>
> I agree here. However, if the RMs also agree. a (RC-)policy
> clarification would be a nice thing to do.
The release policy states: "Packages in main cannot require any software
outside of main for execution or compilation." A recommendation is not a
requirement; I don't believe that unfulfillable Recommends are not
release-critical. If we were to make this release-critical it should
have been done months ago.
We may revisit this post-sarge, depending on the number of packages
affected. Clearly there's a difference between a package in main
recommending a package in non-free (where the non-free package would be
selected by default) and a package in main recommending a non-existing
package (which tends to be treated as essentially a no-op by current
generations of package management tools).
Note that I'm not saying that unfulfillable Recommends aren't a bug.
Cheers,
--
Colin Watson [cjwatson@debian.org]
Reply to: