Where is the FC3 kernel Source?

Sam Williams sam-williams at insightbb.com
Thu Nov 25 09:03:48 UTC 2004


On Wed, 2004-11-24 at 12:53 +0100, Alexander Dalloz wrote:
> Am Mi, den 24.11.2004 schrieb Sam Williams um 7:31:
> 
> > > The FC3 release notes will tell you :)
> > > 
> > > http://fedora.redhat.com/docs/release-notes/fc3/x86/

> We all have short time. And death is the end.
> I can give you my paypal account and if transfer me a small part of the
> multi-million dollar I will read the release notes for you ;)
> Sorry, couldn't resist. In fact, Fedora is for free, this community
> based help forum is free, so I think a bit can be expected by the users.
> And as Fedora Core is developing fast (short lifetime cycles) it is very
> basic to read about the changes. Those are a mentioned / explained
> within the release notes.

Sorry Alex, in order to transfer to your paypal account I would first
need to have a Purchase Order (PO) generated. Sorry I don't make the
rules. Since this takes way too much time its not practical. No offense!

Now everyone has pointed out the obvious.... oooooh why would you use
Fedora Core for this hypercritical project. Its too immature, its
bleeding-edge etc, etc, etc.... Pull up a chair boys and girls.....
Allow me to explain.....

The environment that I'm architecting consists of multi-hardware
architectures... Opterons, x86, and 970-ppc based platforms.  When I
began working with FC2 it was the only OS that claimed to support all
three architectures. Generally it worked great for x86 and  opteron, but
I did find  the yum and apt amd64 repositories stopped working months
ago for xorg. I never did get around to trying it for 970-PPC. On 970-
PPC I continued  using an alpha YellowDog,  released prior to FC2,
because of all the great kernel work being submitted back specifically
into the 64 bit YellowDog kernels by Benjamin Herrenschmidt (sp?).
Thanks Ben!  Nevertheless, I needed the multi-architecture support which
ruled out RHEL, RH9, or FC1.  I also needed an environment which
supported the 2.6 kernel. You see these are either 2 or 4 way multi-
processor machines. I  needed the taskset capability that Robert Love
among others have been working on to  bind processes to dedicated
processors.  I couldn't afford to have processes migrate from processor
to processor at random times or during context switches. The 2.6
requirements meant that I couldn't use the only other multi-architecture
offering at the time,  SuSE Enterprise 8.x (sorry don't remember the
exact number).  Finally, because of the quad graphic cards I used, I was
forced into a redhat/FC  based solution because of the way the vendor
had developed the XFree86/xorg graphics drivers. The drivers were
developed for RH9 and did require a considerable amount of tweaking
under FC.

Anyway, things have changed. SuSE has a very stable offering on all the
aforementioned architectures, it has the purchasable support for those
hard to reach problems, and now offers 2.6 kernel support. I really
didn't want to move into that direction, but probably will. The single
processor machines that I've been testing  recently seem to behave fine
with a minimum amount of problem. The only thing missing from SuSe9.2 or
Professional is the gnome2.8 support which I really like, but this can
be handled with garnome or jhbuild on the control consoles.  It is still
early enough in the project development cycle that work can be easily
transferred allowing for those few distribution dependencies that need
to be addressed. We are coming up on a new fiscal cycle so I can make
sure that I compensate on next years project budget to handle any
increased support cost due to system criticality, although SuSe in all
its forms will now work without purchasing paid support unless
absolutely deemed necessary. Finally, because I'm switching to the
Matrox QID Quad graphics card, I will most definitely use the XiGraphics
X-server instead of xorg because they have been working closely with
Matrox and say  there now exists fully-functional support for this card
under their server for SuSE. I've been doing some soul-searching on this
issue anyway, because they said they won't develop their X-server on
Fedora because Fedora is too unstable.

So you see not all the world is a nail to sadly paraphase Maslow. There
are many details that forced me into trying Fedora Core to solve my
problems. If did  work, but  sometimes at great pain. Most importantly
FC, did  allow me to produce some early proof of concept demonstrations
for our customers which helped to clench several sub-system contracts.
They were impressed with what they saw and the project schedule now has
enough flexibility to allow the change to a new distribution.  It really
isn't much of a loss since with any prototype you always plan on
throwing away one, so this change in course isn't completely unexpected.

So I would agree Fedora Core is too unstable for this type of enterprise
project. Intially, it was required but there are now alternatives that
will  make more sense. And now you know the rest of the story..........

Cheers!

-- 

Sam Williams                                           samurai at acm.org
**********************************************************************
        Welcome to Rivendell Mr. Anderson.......
                                       Machinae Destiny
**********************************************************************




More information about the fedora-list mailing list