[Rpm-maint] various bug fixes (patches)

Ralf Corsepius rc040203 at freenet.de
Thu Nov 15 08:41:26 UTC 2007


On Thu, 2007-11-15 at 10:05 +0200, Panu Matilainen wrote:
> On Wed, 14 Nov 2007, Ralf Corsepius wrote:
> > On Wed, 2007-11-14 at 08:54 -0500, seth vidal wrote:
> >> On Wed, 2007-11-14 at 13:42 +0200, Panu Matilainen wrote:
> >>
> >>> What I'd suggest is:
[...]
> > Also I would like to see some detailed documentation on this feature.
> 
> To put it shortly, this "feature" adds a couple of optional, purely 
> informational tags that applications are free to ignore or use as they see 
> fit, and query switches to display the information. If eg yum doesn't want 
> to implement soft dependencies, all you need to do is ... nothing, nothing 
> at all. The support for optionally installing weak dependencies could be 
> implemented entirely as a yum plugin.
> 
> > Intentionally (non-hostile), devel's advocate questions:
> > - Does any application need this feature?
> 
> Need is speculative 
Erm, no. Implementation should be need/demand driven, not conversely.

Of cause, creating demand by adding features is a common principle of
marketing ("Only products from brand X have feature Y", ... I guess, I
am not the only person to have quite a number of gadgets around being
equipped with some useless ballast "feature Y")

> Enhances and supplements are essentially reverse (soft) requires:
> Package A having "Enhances: B" is the equivalent of package B having 
> "Recommends: A". This is especially useful in situations where you for 
> whatever reason don't have control over the other package. An example: you 
> could have "Enhances: flash-plugin" in nspluginwrapper, and when a user 
> installs the proprietory flash-plugin from Adobe's site, the depsolver 
> would notice that nspluginwrapper should probably be installed with it (or 
> at least notify the user about it). Supplements are the same but treated 
> with same "priority" as suggests.
> 
> The average user would probably be quite happy with the above setting. The 
> people screaming about dependency bloat could turn off installation of any 
> soft dependencies in depsolver config. And the "I want my 
> Everything-install!" folks could turn on installation of all soft 
> dependencies...

Well, I would expect the same folks starting to complain on
"Recommends/Suggest" whenever they find a piece of software doesn't work
because of "missing plugins/add-ons" (Why is X a Recommends and not a
Requires?") or when trying to mix both strategies ("I need 'plugin X for
package Y', but don't want to have 'package series Z' installed.").

I.e. I don't see how this feature solves a problem, I feel it is
shifting around problem (and probably is introducing new ones).

> Again, the above is just one interpretation of soft dependencies, distros 
> and depsolvers are free to do whatever they see fit (including ignore) 
> with the new extra information. Rpm does not enforce *anything* regarding 
> soft dependencies because, by the very definition, the are *soft* 
> dependencies and a missing soft dependency can never be an error of any 
> kind.
> 
> > ATM, I am not really convinced it is _really_ useful.
> 
> Well, in several occasions in my past life, I would've killed to have soft 
> dependency support in rpm and related tools. And clearly SUSE and 
> Mandriva are finding it useful (not to mention Debian and derivates :)

OK, then the Mandriva and SuSE folks probably will be able to elaborate
how they are using it in their distros and their installers. 

Ralf





More information about the Rpm-maint mailing list