On Mon, 2010-11-15 at 23:22 -0500, seth vidal wrote:
On Mon, 2010-11-15 at 20:54 -0500, Jon Masters wrote:
> On Mon, 2010-11-15 at 20:34 -0500, Bill Nottingham wrote:
> > Jesse Keating (jkeating(a)redhat.com) said:
> > > > Do you think it's appropriate to keep it this way or should I
> > > > The reason why I do it this way is, that pungi doesn't pull
> > > > comps packages in the media kit. Any idea how to work this around
> > > > without having an extra script?
> > >
> > > To date, all spins have been either live media, or a virt image that is
> > > quite similar. However you mention using pungi, which means you're
> > > making a more traditional install media, as opposed to live media.
> > > Basically you're just adjusting what packages are or are not on the
> > > media we generate as part of a Fedora release?
> > ... I'm confused why you'd do this as well. How many people install
> > servers from *media*? (That being said, it would be interesting to
> > benchmark an image-based installation of @core, but given that anaconda
> > won't run on that...)
> I always install my own servers from "media". Often it's an ISO given
> a VM, but it is always from an ISO. Never a pre-packaged image, etc.
No - I think he meant a network install.
Actually, I still mostly do non-network installs (even on virtual
machines). I don't think I'm alone, especially in the Fedora space. I
*do* think it's far more common to do network installs the larger you
get, but for e.g. a home VM host box or a few build boxes like I have at
home, I generally download the media and do it from that. That also has
the advantage of not requiring another box to serve the install images
that is local (Fedora installs over the public Internet is painful,
better than it used to be, but still fails too often in my usage).
Cases where I don't do local installs are:
1). Poking with creating virt. instances for e.g. distcc use.
2). Testing anaconda or install-time stuff with boot.iso.