On Dec 19, 2007, at 1:18 PM, Per Øyvind Karlsen wrote:

>> Since today, Mandriva is using rpm.org.
>>
>> So to continue to fully test and work on rpm5.org, I am looking for a
>> distribution which do plan to use rpm5.
>>
>> This surelly mean I'll contribute to it if I find some rpm missing  
>> to me.
>> (Notice I am a mandriva contributer already, I maintain lot of perl
>> packages, rsync, postgresql, some rpm apps, and a lot of others  
>> stuff).
>>
>> So, what distribution do you have to suggest to me ? Feel free to  
>> promote
>> your own distro ! :)
> I'm on the same look as well, but since I'm such stubborn and too  
> comfortable
> with much things technically, I'm on the look for some distro with  
> either
> influence by Mandriva or open for it. Also implied is great freedom  
> to do
> stuff in distro. ;)
> My background is roughly the same as Nanar's :)
>
> ps: also open to any new projects! Feel free to get in touch off- 
> list as
> well :)
>

One of the ideas proposed back in June by jasonc was to set up
a small "reference" distribution. Say, all the packages necessary
to run /bin/sh in a chroot.

The rationale for doing so is to examine (and change) the
manner in which a linux distro is bootstrapped.

There are a number of flaws in the "traditional" Red Hat
(circa 1997) minimal boostrapping package set these days
that I know about.

The list includes:

1) its not clear how to lookup user/group, either inside
or outside the chroot, without functional glibc. parsing
/etc/passwd directly in rpmio/ugid.c is one answer,
patch in OWL linux for years.

2) glibc (wrongly imho) uses a statically linked helper
that tries to restart ssh. While I understand why that
is being done, I think there are far simpler means
to accomplish the same goal (insuring that a connection
to a remote colo is not lost if/when name services in glibc
change incompatibly). lua, or a trigger, is better design.

3) basesystem is just a useless pkg these days. AFAIK,
the original intent was to use the basesystem version as the
distro version, but that rule is imperfectly followed many years now.

4) filesystem after setup pkg is odd, and the %ghost entries
in the setup package are an accident waiting to happen (%ghost
files do not have a well defined file type is the flaw) that are due
more to Bill Nottingham's dry sense of humor than any real world reason.

5) ownership of directories, particularly for i18n locale files, is
increasingly important. What is actually happening in most
linux distros today is stooopid, but noone has a clue anymore.

I can probally continues with 3-4 additional flaws, those
are just the flaws that instantly come to mind.

And a "reference" distro needs to be achieved, whether
through artificial insemination or adoption, in order to
establish reproducers for the rpmlib install state machine.

73 de Jeff