[pkg-wpa-devel] Re: wifiroamd as a Summer of Code project

[Alexej Davidov]

On Tue, 18 Apr 2006 13:41:18 +0200
Joachim Breitner <nomeata at debian.org> wrote:

> Hi,
> 
> Am Dienstag, den 18.04.2006, 13:17 +0200 schrieb Alexej Davidov:
> > > ...
> > > it should associate with that accesspoint and give control to
> > > ifupdown.
> > [...]
> > In my opinion, wifiroamd should only do roaming, no bringing up of
> > interfaces. We have ifplugd for this purpose and it works perfectly
> > (at least for me). 
> > [...]
> 
> No problem, just insert "indirectly" after "and" in my sentence :-)
Good.

> > > It can be thought of as a ifplugd for wireless devices.
> > 
> > But it shouldn't. ifplugd is already there. No need to reinvent the
> > wheel.
> 
> Sure, if the way wpasupplicant currently signals ifplugd the
> availability of "link" is not too much a hack, and it doesn't seem
> like it, as it works fine, we can stick to the current behaviour.
Good.
> 
> > Maybe the approach you've suggested seems logical, once one has
> > approved of the new ifupdown config for wpasupplicant, which already
> > introduces such an arbitrary distinction between wired (where you
> > can use ifplugd) and wirelese interfaces (where you can't and you
> > have to reinvent it using other means). So my problem with the
> > suggested solution might be more fundamental, as I don't approve of
> > the new config scheme, to begin with.
> 
> I like the old setup too, but when a roaming daemon is available, the
> per-network-encryption stancas of the new scheme make sense, IMHO.
> Without a roadming daemone, although, well, see the list traffic :-)
> 
From what you write, it seems my concerns were unfounded. If I
understand correctly, it still will be possible that wifiroamd is
running as a daemon doing the association according to some config
file, than let ifplugd do the beat detection and call if-up. 

I'm still not convinced that the /etc/network/interfaces approach
is the best way to set encryption parameters. But I want to give it
some more thought before commenting on thisn. Anyway, with the split
of wpasupplicant in a roaming part and an encryption management part,
this point is not relevant to wifiroamd and can be discussed separately.

Regards,
A.