Xen support without reinventing the wheel

[Matthew Palmer]

On Tue, Nov 08, 2005 at 09:19:28AM -0500, Eric S. Johansson wrote:
> 'Forum Post wrote:
> >It is also appeared on "xen-users" list:
> >
> >*[Xen-users] Xen packages availible for Ubuntu and Debian*
> >http://tinyurl.com/e2cuf
> >
> >Interestingly, all the links to READMEs in these post are broken.
> >Does the SoC really finished? Or is it really suspended?
> 
> It's dead Jim.
> 
> we need two things.  First is a production method of generating xen 
> kernels from the standard released kernels.

kernel-package has supported building Xen kernels for a while now.  The
standard way is to build a kernel-patch-xen from the Xen sources, and then
apply that to the regular Ubuntu-provided kernel source.

> I think probably suffice to take a source deb of the kernel, add the
> appropriate patches and turn on reasonable defaults and generate a new
> source .deb for a dom) and domU kernel

Well, yeah...

> then we would need a .deb for xen-tools which would be all of the 
> surrounding components for xen management.

The existing Debian packages for Xen provide this side of the coin quite
well, and I don't think there's any need to radically change things here.

> if folks want to try and solve this problem, I would be glad to 
> contribute what knowledge I have about xen and find some shortcomings.  ;-)

My employer is getting deeply involved in Xen, so I'll probably be thrown
onto the problem if it gets too bad.  Our current Xen infrastructure is
built around the stock Debian 2.0.6 Xen packages, but with "custom" built
kernel packages (for hardware support reasons).

> I would suggest discussing this activity on a transient mailing list 
> (quicktopics.com which is run by a friend of mine comes to mind) so we 
> wouldn't annoy the people here.

No, keep it here.  It's Ubuntu development, and the idea of a "transient"
mailing list just gives me the willies.  It takes a good couple of days to
get any sensible number of people subscribed, and momentum isn't something
you can just pause.  Also, MLs get good when they get wide-spread publicity,
and a transient list would have no permanence and hence no publicity.  Also,
a major benefit of mailing lists is their archives; I'd imagine that a
transient list would have it's archives blotted after a while (and even if
they weren't, if the list itself is gone, then if the question isn't
answered in the archives, where do you go?).

> are we ready to move forward?

To where?  My suggestion would be to look at the existing Debian stuff, and
if it is deficient in some area, identify the deficiency and work out some
way to fix it.  Then fix it.

- Matt