Fedora Dev Spin/Strain

[Jane Dogalt]

I recently outlined the simplest example of my ongoing pursuit of
trivially self-replicating (and modifying) livecd distributions.  I.e.
the goal of having an fc8 installable blu-ray(whatever) which also
behaves as a livecd, which can regenerate itself with arbitrary
modifications via simple 1-click/simple user interface.

Going further down this line of thought, takes me to this outline for
the ultimate fedora8 development distribution livedvd image-

In addition to the above features, the livedvd environment should-

1) have a full copy of the source repo (src rpms).

1B) It would be better to also have a full copy of the
cvs/svn/git/whatever repository that builds the srcrpms repo from
network upstream sources (presumably via 1click/commandline
non-interactive interface)

2) have a version of pungi or whatever that non-interactively inputs
(1)
and outputs the full binary repository.

2B) with an option for using qemu-ppc to generate even the other binary
repositories (after quite a bit of unattended gratuitous power
consumption)

3) a tool like pilgrim or what I'm working on that inputs (2) and
outputs an installable livedvd image.  I.e. the environment you are
using to build this new modified version.

4) now here is where I'm really asking for a lot:  Now that you have
every line of source code on this self replicating dvd, I would also
like to see a database where literally every bit(/line of source code)
on the dvd is associated with a maintainer/upstream-maintainer.  Such
that at any point any developer (or bright kid) with sufficient skill,
can dig into any bug they find, down to the relevent lines of source,
and then trivially get that information and/or patch to the correct
upstream owner.

5) As I've been harping all for some time, you also want the feature to
virtually run new output livedvds (in qemu/etc).

6) what (5) is really for, is to facilitate fully automated regression
tests.  In addition to being self-reproducable and self-debuggable,
they should be viciously self-testing as well.

7) the reason that all of this hasn't been done so far, is (a) its
quite
complicated tying a lot of this together.  but imho this is just a
level
of maturity that fedora just hasn't achieved yet, but its getting
close.
and (b) on current hardware this is maybe a runtime of a week or more. 
But this is where you also want to bring in the tools to trivially
allow this livedvd to boot an entire cluster of xbox's on a lan, and
give you trivial 1-button interface to rebuild/reproduce all in such a
distributed system.  Likewise, this may just take 30 minutes to run on
a middle-class laptop 10 years from now, and can be the official new
'order a pizza' step in the compile your own kernel(&now distro) howto.

-dmc/jdog


 
____________________________________________________________________________________
Have a burning question?  
Go to www.Answers.yahoo.com and get answers from real people who know.