Re: Thoughts on network detection and configuration on Debian
Enrico Zini wrote:
> Debian seems not to have a masterplan to define its behaviour
> for a system that should react to a changing configuration.
A master plan (i.e., policy) is clearly needed.
> Maybe it's time to design a tool that gathers all notification
> of the changes that happen in the system and provides the user
> with a single point to program the logic to react to them, and
> standardize it in such a way that other parts of Debian can
> interact with it?
A master plan does not entail having a master tool, however.
It has been suggested that at least three tools are
required, but I think that there should be a separate
low-level configurator for each subsystem. If we want to
give the user a unified interface to all the configurators
then that can also be provided.
What is needed in policy is a specification of how configurators
are to be notified of configuration changes (or alternatively,
of how configurators should poll for config changes). The
mechanism needs to take into account the fact that there are
dependencies and conflicts in reconfiguration. Not a trivial
problem.
As a general comment, let me just say that Debian needs to act
on this. I was just playing with Windows 2000 for the first
time and I was very impressed with the latter's plug-and-play
behavior. What impressed me more than anything was the fact
that I was able to move my W2K hard drive to another machine
with a different processor and different installed hardware,
and W2K booted without any problem. It simply detected the
new hardware and hummed along. Debian must aim to be equally
well behaved, IMHO.
--
Thomas Hood
Reply to: