Shtt!!

[Michal Jaegermann]

On Tue, Dec 02, 2008 at 05:22:18PM -0500, Jeremy Katz wrote:
> Then instead of using the mirror returned by MirrorManager,
> you'd get the very very local mirror.  Also, it's nice for things like
> an installfest where you might not even have a public network
> connection[1].

In situations like that it is the best to have a local, _updated_,
repository and they you can give all 'repo=...' and 'stage2=...'
parameters, which include method specification, to anaconda.

If you are booting over PXE then all these extra parameters can be
easily put into a boot menu so later you do not have to bother with
all that stuff.  I was already installing F10 from a setup like
that, with a repo accessed over NFS, so this is not just
theoretical.  When you are preparing other boot media then you
should be able to do that as well.

> I'm not 100% buying into it either -- otherwise, I would have worked on
> actually pushing it more.  But it does seem to come up with some level
> of frequency, and unlike some of the things in that bucket, I can
> actually see some of the cases where it's useful

It could possibly be useful and simplify things for those who set
that up but then anaconda would have to have some automatic
preference for such local repo.  It already does?

> [1] Although I suspect that these days, most installfests *do* have a
> public network connection.

If that happens to a be some kind of ADSL, or worse, and not T1 then
I doubt if that helps more than hinders.  Even with T1 I would
prefer to keep traffic "local" as far as possible.

    Michal