[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Guessing gateway by ip-address...

"Bernhard R. Link" <brl@pcpool00.mathematik.uni-freiburg.de> writes:

> * Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de> [031006 17:55]:
> > > Ok, that explains the behaviour. Now my confusion is limited to the
> > > question why anyone might think an IP-address without the dns and/or 
> > > gateway address is of any value.
> > 
> > Given an IP, say I would guess the gateway to be
> >, then netmask to be /24 and the DNS to be too.
> The netmask might be quite usual, but I'd see the gateway as a 50% bet.
> At least I was told there are nets with .1 instead of .254, which I find 
> in my environment exlusively.

Ok, .1 and .254. Two guesses that can be tried by configuring them and
sending a ping out. :) It should certainly be tried in some way and if
no working one is found the question should be asked despite its

Or, looking at it the other way, the autodetect should preseed the
answeres (if it found any) and lower the priority.
> And if I remember correctly (Neighter have the computers to test
> nor the checked out d-i source near currently) the dns-servers was the
> last thing it asked (i.e. has the lowest priority) while it seems to
> be the wildest guess.
> > DNS being the wildest guess here. I would prefer a multiline formular
> > there. One field per question and while you change a field the other,
> > not yet edited fields, would change according tothe guessing
> > algorithm.
> > 
> > So normaly one would just enter in the IP field and skip
> > ahead to "continue".
> > 
> > Any volunteres to add a new widget to (c)debconf?
> I think the old way was good enough, in asking IP and then give those
> guessed defaults pre-filed in the next questions. (Thus easy to press
> Enter if they are correct guess. And also normaly more easily changed
> to the correct values than typing the whole at once).

Conflicts with the goal of some of us to minimize the number of corns
we have to place on the return key to get a chicken to install debian.

> > > [1]: While it finaly configured the network properly, it never
> > > exited without error, as it stumbled over doing things twice.
> > 
> > That should never happen :) Bad, bad, bad. Happy bug hunting.
> Well, that looks like an easy one. Will put it on my todo list to
> look what it was exactly. But the point is that things like this
> will always happen. There is no easy way to ensure all scripts
> are idempotent and none script will ever fail by returncode when
> it did all it had to do corectly. (When it shall also report
> real errors as such)

I allways expect the worst and hope for the best :)


Reply to: