[CentOS] build Centos source/rpms from scratch...

R P Herrold herrold at owlriver.com
Sat Aug 19 21:07:42 UTC 2006


On Sat, 19 Aug 2006, Jim Perrin wrote:

> It's been stated a couple times on the list that there's no 
> interest in helping people rebuild the entire distribution 
> from scratch.
...
>   beyond that, I'd say you're on your own. Too much 
> rebuilding dilutes the gene pool.

Jim states it humorously, but a whole swarm of issues with 
emerge during a rebuild effort:
 	- Will package building be done in an 'as found' build 
environment, or in a chroot
 	- As non-root, or as root
 	- Will the build of the packages be done in a minimal 
build chroot, a defined chroot, a fully populated build chroot
 	- Will the build chroot be pristine every time, or are 
shortcuts tolerated
 	- How will non-explicit BuildReq's to be handled
 	- What arches are targets
 	- What installation methods will be supported
 	- Is it a goal to maintain ABI interop with the 
upstream PNAEVL, or is drift tolerated
 	- What workarounds will be used to solve reaching a 
ABI identicality when the upstream has build with a tool (such 
as a GCC variant) not released
 	- What post-testing processes will be used to measure 
confirmance
 	- Is it going to solve the trademark issues, or ignore 
them
 	- Will all package versions match the upstream 
identically, or will 'creep' of later versions and fixes be 
accepted
 	- How will updates be handled

We have not even touched on issues surrounding binary and 
source code distribution, support, documentation, mailing 
lists, bug trackers, community building, and the like.

Thinking of material alternative efforts similar to CentOS', 
David Parsley worked and published his solution first, has as 
to RHAS21; John Morris later with his writeup as to RHEL 3; 
and many others as well within the context of the 'rebuilds' 
of RH Enterprise sources derived distributions.  Each has made 
differing choices, and to greater and lesser extents, 
documented their efforts.  David has retired Tao into CentOS 
recently, as it was so demanding; others as well have flagged 
in their 'freshness' over time.

A strong aspect of CentOS is the ability to attract and retain 
committed volunteers, and really become the 'community' 
solution for the whole range of 'being an Enterprise-grade 
Distribution' -- it is far more than just compiling packages 
so that they can be wedged into an installable format.

Here, as with the IRC channel's guidelines, well asked 
questions may draw a meaningful response from a CentOS 
'insider'; general ones will (and I argue, _should_) not.

Solve about the Cost/Benefit equation from the standpoint of a 
CentOS volunteer: Asking how to make CentOS weaker (as by a 
random call not showing the slightest bit of advance research 
and being a distractor from more pressing tasks) is a question 
thoughtlessly asked.  Within the range of proper replies: 
devnull it and move on.  Another couple approaches from the 
core group also have been offered.

Thanks, CentOS core group.  You know who you are.

- Russ Herrold



More information about the CentOS mailing list