NM and NM applet quality status update

[Dan Williams <dcbw redhat com>]

Hi,

So I didn't answer a couple mails about the quality stuff yesterday because I 
was working through it.  A few points:

1) The applet was using the quality information from the active _network_, not 
the active _device_.  I fixed that.  The problem is that we can get much better 
quality statistics for the device (using SIOCGIWSTATS) than we can from the 
individual networks (which is from the scan information).  Some cards don't 
provide quality data from scans, or don't fill in the right information.  This 
should fix some of the quality issues people were seeing.

2) The NM method of determining quality in nm_wireless_qual_to_percent() was 
busted as Bill has already pointed out, I've tried to fix that somewhat.  
Instead of using Bill's method directly, I've re-worked that code (still using 
Bill's method for some of it) to be conformant to my understanding of the WEXT 
quality spec, which I've actually posted my understanding of to netdev list:

http://oss.sgi.com/projects/netdev/archive/2005-02/msg00149.html

If my understanding is in error, then I'll fix that assumption in the code in 
NetworkManagerWireless.c.  Otherwise, we get to fix the drivers.  I've decided 
that its not worth working around and special-casing driver problems here, if 
its busted for particular cards, people will have more incentive to get it fixed 
:)

This may necessitate updating the ipw2200 patch, Bill, to bring it into line 
with WEXT API.

I'll be committing this code soon, once I figure out a few more things.

Dan