(Very) bogus package dependencies

[Stephen Montgomery-Smith]

Paul Schmehl wrote:
> --On Monday, December 10, 2007 16:33:20 +0200 Andriy Gapon 
> <avg at icyb.net.ua> wrote:
>>
>>> From a small research it seems that the only thing needed from cdrtools
>> is isoinfo utility which gets called in FreeBSD-specific code
>> (hald/freebsd/probing/probe-volume.c) like follows:
>> isoinfo -p -i %s
>> And it seems that its only usage is to detect presence of directories
>> named 'VIDEO_TS|VCD|SVCD', so that properties like
>> volume.disc.is_videodvd could be set.
>>
>> Maybe there is a way to write code for this functionality that could be
>> included into hal source code or as a port patch, so that hal doesn't
>> have to depend on cdrtools.
> 
> While I have no objections to this particular  suggestion, my question 
> would be - where do you stop?  You could easily do this for hundreds (if 
> not thousands of ports) that depend upon some other port because of one 
> piece of code.
> 
> In general. port maintainers follow the guidelines of the software 
> developer.  If the developer states that the software depends upon 
> cdrtools, then the maintainer is going to include that dependency in the 
> port.  Many of us don't have sufficient skill to audit code and 
> determine where a dependency could be replaced by some additional code.
> 
> So, while this might make sense in isolated cases, I don't think it 
> scales well.  Furthermore, modern machines generally have enough disc 
> space that the addition of a few "unused" ports to include necessary 
> code is a small price to pay to distribute the load of providing ports 
> over a larger population of volunteers.  (And yes, I know not everyone 
> has a modern machine or large discs to work with.)

I agree.  I think that in this case, cdrtools is a legitimate 
dependency.  The fault (if any) lies not with FreeBSD, but either with 
the writers of hal, or the writers of xorg who made hal a dependency.