Re: Apples and snakes

[Jeremy Katz <katzj redhat com>]

On Wed, 2005-09-21 at 16:02 +0800, John Summerfied wrote:
> Jeremy Katz wrote:
> > On Wed, 2005-09-21 at 11:32 +0800, John Summerfied wrote:
> >>Jeremy Katz wrote:
> >>>On Wed, 2005-09-21 at 08:54 +0800, John Summerfied wrote:
> >>>>Ah yes, but my proposal allows booting a standard 
> >>>>just-bought-from-redhat CD and booting it, automatically fetching a ks
> >>>>file if Anaconda can find one, without one needing to specify its 
> >>>>existance in any syslinux, yaboot or other boot tool's config file.
> >>>
> >>>Unfortunately, this would not be the desired behavior of other people.
> >>>So as it stands, we're going to stick to requiring the kickstart config
> >>>to be specified.
> >>
> >>What problems do you see?
> > 
> > 
> > A lot of PXE setups are set up such that the filename of pxelinux is
> > always given out.  So by always following that, we'd end up grabbing a
> > garbage file which could then end up having less than desired effects.
> 
> I've already dealt with that one: Anaconda should recognize a pxelinux 
> file (or other binary for that matter) and silently ignore ot.

The silent ignoring is actually something that is causing lots of grief
-- people want to know about the failure.  While looking on tty3 is
great for local installs, it's less useful on hardware that doesn't have
a physical console.

> > While the class based and dhcp identifier stuff is nice, I'd venture a
> > guess that most people aren't actually using it and I'm sure there are a
> > number of commonly used dhcp servers out there that really don't support
> > it.
> 
> They don't _have_ to use it if Anaconda supports it. If Anaconda 
> supports it, they _can_ use it.
> 
> AFAIK RH/Fedora only documents network installs from RH/Fedora servers, 
> and those have ISC DHCPD 3 so there's no problem there. DHCPD 3 also 
> ships with Debian, another considerable base of Linux users, so those 
> don't have a problem either.

It's nearly impossible to document things we don't ship.  But that
hardly means they're the only things that a) people use or b) we
support.  If I only had to deal with homogeneous environments, my life
would definitely be much simpler :-)

Jeremy