[pkg-wpa-devel] Non-roaming usage of the daemon: Radio kill switch

[Felix Homann]

On Saturday 08 April 2006 16:42, Reinhard Tartler wrote:
> I think that the existing README.Debian, README.modes and Debian.NEWS
> files make very clear that there is now a new behavior 

Yes.

> and explain how 
> that can be used in order to implement roaming.

I don't have a roaming issue here, I want a nicely working wlan button. But 
even the documentation on how to implement roaming with the new setup is not 
very satisfying. 

> If youare unhappy with them, I'm happy to read your suggestions how to
> reword them to make them even more clear. What kind of additional
> information did you expect? and where?

I've just prepared a seperate mail for this.

>
> > It just doesn't work as expected!
>
> There seems to be a shift in expectations.

Hmm, when something is declared deprecated I expect that the new concept can 
at least reproduce the former functionality.

> > (First, let me correct myself, unfortunately I've mixed up some things: I
> > used to use wpa_supplicant in daemon mode and *ifplugd* for watching the
> > interface.)
>
> I think this is quite braindead.

Please, don't get insulting!

> Nowhere in the old package it was 
> advertised that wpasupplicant would work with ifplugd at all. It rather
> happens to work with that, and that specific undocumented behavior 

Well, would you even call the kernel's ability to run wpa_supplicant 
a "specific undocumented behavior"? It is the other way round: ifplugd is 
intended to work with e.g. wpa_supplicant exactly the way it does/did. You 
simply don't want this behavior anymore.

> Now we changed the situation,

Yes, and things got broken. As you can tell from other messages on the list 
using ifplugd together with a "mode 3" wpa_supplicant is very common. As long 
as the new approach can't reproduce it's functionality the init script 
approach should not be deprecated.

Keep in mind that the wpa_supplicant/ifplugd solution is not only useful for 
roaming but even for the simple (?) task of having a nicely behaving wlan 
button.

> > At first sight it didn't seem to work at all: The documentation doesn't
> > tell that you have to make the script executable.
>
> well, I'm happy you figured that out. it is a script, after all.

It should be mentioned nevertheless. It's one more unneccessary hurdle to 
take.

> sure. you can customize the action script to your needs

In which way, how? It is rather undocumented!

> and perhaps 
> lower the timeout. 

Where? (I actually know where, but it's not documented). I guess the timeout 
is there for a reason. Would it have any unwanted side effects to lower the 
timeout to 0? I don't think a user should mess with ifupdown.sh.

> If you cannot guarantee that you have a working location at system
> bootup, don't set 'auto wlan0' in /etc/network/interfaces. 

Sure, but then again I would not have a nicely working wlan button! I would 
have to login to ifup the interface.

After all, the new approach seems to be very close to what the old 
wpa_supplicant together with ifplugd did. If it 
wasn't for the boot/timeout issue it would seem like a working replacement 
(well, I haven't  tested suspend so far with the new approach..).

Kind regards,

Felix