Quoting Reinhard Tartler (2014-11-23 02:57:33) > On Thu, Nov 20, 2014 at 9:02 PM, Jonas Smedegaard <dr@jones.dk> wrote: >> Quoting Reinhard Tartler (2014-11-20 21:45:56) >>> On Nov 20, 2014 3:01 PM, "Jonas Smedegaard" <[1]dr@jones.dk> wrote: >>>> Quoting Andreas Cadhalpun (2014-11-20 17:09:49) >>>>> On 19.11.2014 13:09, Jonas Smedegaard wrote: >>>>>> Possibly we can simplify even further: >>>>>> >>>>>> * Have package libavcodec-extra-NN provide virtual >>>>>> libavcodec-extra (i.e. non-versioned name of itself) >>>>>> * Let GPLv2 packages conflict against libavcodec-extra (i.e. not >>>>>> replace but complement existing suggests/recommends/depends). >>>>>> >>>>>> How does that sound? >>>>> >>>>> This sounds good, except that the virtual package needs another >>>>> name, because libavcodec-extra is actually a real package [1]. >>>> >>>> I don't see a problem in conflicting both with virtual provisions >>>> of the package and the real package - as the latter seems to me to >>>> provide nothing beyond pulling in latest of those same virtual >>>> provisions. >>> >>> I'm pretty sure that this approach makes it impossible to have a >>> versioned dependency on libavcodec-extra. Not sure if there is an >>> actual need for this, though. >> >> Now you get me curious: How would providing a virtual package make it >> any more difficult to depend versioned on a package? > > > AFAIUI, if you have libavcodec-extra both virtual and real, then > application packages can no longer declare: > > Depends: libavcodec-extra (>> 6:11) > > And reliably expect that the version constraint is obeyed. Or am I > missing something? Until recently, declaring a versioned dependency against a package both real and virtual meant that only the real package was considered, by the logic that only real packages can satisfy a version constraint. Recently dpkg and APT has grown support for versioned virtual dependency, I believe. I don't know how that works, but I am confident that both old and new behavior satisfies this use case: limiting to known good package (possibly then limiting too much, but that's for the package maintainer to then adjust). >> Is it more difficult to depend versioned on e.g. apache2, which >> provides httpd and httpd-cgi? > > I believe it is also not possible to reliably declare: > > Depend: httpd (>> 2.0) > > Reinhard > > PS: Wouldn't that be great to have ensure to have a Web 2.0 service > installed? ;-) Well, that might actually be possible now :-) - Jonas -- * Jonas Smedegaard - idealist & Internet-arkitekt * Tlf.: +45 40843136 Website: http://dr.jones.dk/ [x] quote me freely [ ] ask before reusing [ ] keep private
Attachment:
signature.asc
Description: signature