[Rpm-metadata] groups-sample.xml and groups.dtd are misleading a bit

[Hans-Peter Jansen]

Hi Seth et.al.,

today I tackled the yum group setup on my local SuSE 9.3 systems.
First, I've created a script to convert the SuSE .sel(ection) files into 
the yum preferred xml format, based on your nice yumgengroups.py.
During that course, I came across an inconsistency in 
http://linux.duke.edu/projects/metadata/samples/groups-sample.xml and
http://linux.duke.edu/projects/metadata/dtd/groups.dtd: the <groupid> 
tag. Simply using <id>, like yumgengroups.py does, works as expected.
Would you throw this script into yum/download/misc?

BTW: the createrepo man page raised the question, in which repo the 
group definition file should/can go? All repos, the first, any ...

Could you elaborate a bit, how the different type definitions of 
groupreq and packagereq tangent yum?

Your script makes sure, that only real rpms will pass. SuSE on the other 
hand, uses heavily the feature to just ignore unresolvable packages, 
e.g. they have many *-32bit packages in their groups: they're 
inexistent in the ia32 world, but came to existence on x86_64..

AFAICS, yum does ignore unresolvables silently, too. Can I depend on 
this feature?

Thanks to this stuff, I successfully installed a SuSE 9.3 default setup 
into an installroot directory with yum today. Great, as this is the 
first major building block to a nice and handy diskless environment.. 
Note: in order to avoid millions of "user/group does not exist" 
warnings (which aren't restorable easily either), I found it necessary 
to install the glibc with rpm beforehand, together with some 
dummy /etc/{passwd,group} files.

Thanks to everybody involved!

Pete