Re: Slight flame at NM

[Dan Williams <dcbw redhat com>]

On Sat, 26 Mar 2005, Timothy Murphy wrote:
> I spent a few hours trying to get NM working 
> with my classic Orinoco 802.11b Gold PCMCIA card, 
> on my Fedora-3 kernel-2.6.11 laptop.

I happen to have one too, but it has issues with associating with weak access 
points in the presence of stonger ones, which firmware upgrades don't seem to 
help.

> (1) Don't you think you should warn people to test their cards
> in this way _before_ going to the trouble
> of installing NM, and trying to get it to work?

Perhaps...

> (2) If iwlist can work out that the card will not work,
> surely NM must know this,
> so why does it not politely exit with a message to this effect?

Surely this is a better option than #1, some notification that the card won't 
work as expected.  However, NM will never "politely exit" at any point, since 
its a daemon and sticks around for hot plug events.  The more appropriate 
scenario is this:

1) you boot up with the Orinoco card inserted for the first time
2) NM notices the orinoco doesn't support scanning, signals NMI
3) NMI throws up the "won't work because it doesn't scan" dialog, including a 
"don't notify me again" checkbox
4) When you click OK, NMI sends NM a message if the checkbox was clicked

Something even better would be to have this type of status in the applet on the 
panel, but I'm unsure how best to display that and I'd have to talk to Seth or 
Brian (Interaction people) on what's appropriate given the gnome HIG.

But you're right, this should be taken care of in a better manner.  In some way, 
the driver is to be considered "broken".  I've personally spent 3 days creating 
the patch set that would merge upstream orinoco into the kernel, but it turns 
out it was the wrong CVS branch.  David Hinds (who I believe is the maintainer 
of the drivers, at orinoco.sf.net) has started doing merges too, but hasn't done 
the scanning code because its one of the larger chunks and adds in wireless 
"spy" support too.

Dan