Re: [Quantian-general] Quantian + Knoppix DVD
Hi Marco,
On 12 August 2005 at 09:36, Marco Caliari wrote:
| On Sun, 17 Jul 2005, Dirk Eddelbuettel wrote:
|
| > Quantian is already setup to support shared work between several people with
| > the Quantian corner on alioth.debian.org -- a hosting site open to everybody
| > and not just Debianers. Right now, it is mostly Marco who contributed by
| > adding to the kernel builds and boot binaries He may even try to help us to
| > get openMosix into the Knoppix 4.0 kernel, who knows.
|
| Hi Dirk.
|
| Yes, I may try to get openMosix into the Knoppix 4.0.
Excellent to hear that -- you'd be clearly the best positioned among us
to do this should Wim not be able to pick up the slack, given how much you've
already done on customising the kernel and boot code.
| But, is it safe?
My reading on the openmosix-general list is inconclusive as the 2.6.* patches
for oM aren't quite ready.
| There are only unofficial patches for 2.6.11. Maybe it is safer to keep
| 2 distinct kernels, a vanilla 2.6.11 (syncrhonized with Knoppix) and the
| 2.4.27-om. Other options:
Agreed! Wim and I discussed that briefly a few months back and concluded
that an update was (then !) a nice-to-have but not strictly necessary as core
Knoppix hadn't moved that much. I haven't seen the not-yet-formally-released
Knoppix 4.0 but I understand it has changed quite a bit (cdrom and dvd
versions, lots of newer software, lots of software from other sources incl
Quantian, ...)
Keeping a solid 2.4.* with a solid oM patch is definitely a good idea as oM
for 2.6.* keeps being described as 'really soon now' but doesn't show up.
| 1) a Quantian version without om
That is a possibility I considered. I'd much prefer Quantian to retain oM, though.
| 2) synchronize with ParallelKnoppix (by the way, I hope ParallelKnoppix
| and ClusterKnoppix can merge a day...)
Agreed. I was in Michael's area this summer, unfortunately he was in the
States! (Hola Michael!)
|
| > I'd be up for trying to coordinate some updates/remasterings to Knoppix 4.0
| > 'lite' and 'full size'. Your idea of starting with the cdrom sized version
| > is good as it requires less hd space, and ram, to remaster. It also has a
| > much shorter 'production cycle' to compress and remaster, ie you can test
| > things faster. Once we're happy with a new lite version, we can go and add
| > stuff to the biggie.
|
| Now there is UnionFS. In Knoppix 4.0 DVD there are two images (KNOPPIX and
| KNOPPIX2) mounted together with unionfs (also to overcome the 2gb
| iso9660 limit, I think).
That is what I read too. The difficulty is to 'split' the filesystem so that
the first image, KNOPPIX, has enough to boot, so one can't just split between
/usr and /usr/lib, say.
| Next Quantian could be made of
|
| 1) a base KNOPPIX image with the most important tools (which tools?...) in
| order to fit a CD.
Exactly! And that is the best way to test / customise / polish the kernel and
oM parts.
| 2) additional images (e.g., KNOPPIX2 for biology, KNOPPIX3 for physics,
| ..., KNOPPIXn for extra - OpenOffice, Gimp, Scribus,...)
Could do, yes. On the other hand: a) other disciplines sometimes profit from
"your" tools so splitting may be bad, and b) the 4gb of Knoppix is twice the
size of our last 0.6.9.3 (!!) so we may not need that much more if we delete
some fluff off Knoppix. That may be harder than I now think, though.
| One could download the CD iso, the DVD iso (containing all the images) or
| the images he wants and burns its own version (given the minirts and the
| KNOPPIX? images, it is quite simple to make the iso).
|
| What do you think about this?
That last 'Quantian as a collection of mergeable images' idea is very
nice. We should try that, time permitting -- but first we need to get a new
0.7.9.* series started once Knoppix 4.0.1 is released.
Also, had you heard about debian-science? New mailing list where a few
related topics are being tossed around. I'll CC this for good measure.
I'm off in an hour for a conference on Seattle (on distributed statistical
computing) where I'll talk about Quantian and oM/MPI/PVM, especially via
higher-level tools from R. I'll post the pdf of my slides when I'm back on
Monday evening.
Best regards, Dirk
--
Statistics: The (futile) attempt to offer certainty about uncertainty.
-- Roger Koenker, 'Dictionary of Received Ideas of Statistics'
Reply to: