Re: ppp using NM

[Marcel Holtmann <marcel holtmann org>]

Hi Dan,

> > And we don't talk about the interference between Bluetooth and Wifi when
> > we keep scanning all the time. Use a spectrum analyzer to see what
> > happens. Bluetooth is designed in a way that makes it quite robust, but
> > it has its limits. Wifi on the hand is not robust at all.
> 
> This is just something people have to suck up.  Until all laptops get
> GPS units or accelerometers and until NM can figure out when you've
> moved the laptop (or at least what the velocity is and for how long, and
> therefore have an idea if you're moving in/out of wifi range), NM is
> going to keep scanning periodically.  Drivers also quite suck here,
> since only now are we beginning to see driver integration between BT and
> WiFi chips to make sure they don't stomp on each other.
> 
> This is a question of user experience, and there is no way NM is going
> to have a "Scan now" button, because that experience just sucks.  If the
> system has the capability for dynamically investigating the environment,
> and the costs aren't that high (as in wifi), then we need to use that
> method to make the user aware of the surroundings.
> 
> NM backs the scan interval off to 2 minutes when nothing has really been
> going on, which is a good compromise between responsiveness of the UI
> and power/interference considerations.

even that is the wrong approach. If you have a fully established
connection with one AP, then the extra scanning only drains battery.
This is totally unneeded and also bad for the user experience. Not
talking about the inference that a scanning might cause to the already
established connection with the AP. Some card firmwares try to cope with
it, but it is still a problem in some cases. Use a spectrum analyzer to
see what actually goes on in the 2.4 GHz band.

If you look at the latest MacOS, you see that it first starts scanning
when you actually press the icon or open a dialog. I think that is good
enough for a "Scan now" button and doesn't interfere with the usability.
There are other cases like directly after resume etc., but the generic
scanning in the background is useless if we have a connection that is
good enough and especially if that connection is in use. An integration
with the upcoming pm_qos framework should be strongly considered.

And this is only your own machine. Think about the other systems that
might be affected by any periodic scanning. The rule of thumb should be
to create as less radio interference as possible.

Regards

Marcel