WiFi power conservation

[John Gilmore]

> >   > Has the following idea already come up?  How about just turning off
> >   > the wlan entirely during suspend, if the machine has reason to
> >   > believe that its contribution to mesh connectivity is negligible?

We can do better, without losing any functionality in school deployments.

First, simply add a GUI control to turn off the WiFi entirely.  This
would save about a third of the power during suspend (and save the
same 0.7 watts during live time too), under the user's control.  This
would be useful whether the user is a kid trying to get their homework
done before the battery runs out; a G1G1 user who isn't accessing the
Internet; someone working on an airplane; or anybody else.

Second, simply add a GUI control to turn off the Mesh part of the WiFi
entirely.  This would save significant power (Mesh is transmitting all
the time; an ordinary WiFi connection isn't).  It would also allow the
ordinary WiFi to go into ordinary 802.11 power saving mode (which
negotiates with the access point so that transmissions from the access
point to the laptop will only occur during pre-negotiated time
windows; the laptop's radio can stop listening the rest of the time).
None of this is fancy OLPC crock-schtupping-hat technology, it's merely
what every other WiFi in the world is doing -- but OLPC isn't.

With those two controls in place, then it's clear to Ohm when more
power can be saved during both automatic and manual suspends:

  *  During automatic suspend, if the mesh is off and we aren't
     associated with an access point, power down the whole radio.  We
     aren't going to send any packets, and nobody's going to send us
     any packets.  Nobody from outside is going to send us anything
     useful, even if we move into range of an access point; we have
     no known IP address.

  *  During manual suspend, if the mesh is off, power down the whole
     radio.  We closed the lid; whether or not we're associated with
     an access point, incoming packets aren't going to wake us up.
     3/4 watt goes away; battery life lengthens by 33%.

Neither of these saves power when meshing.  But both save lots of
power when not meshing.  And all of the complex ideas for
saving power in mesh mode are also available in this scenario --
if and when anyone designs, implements, and debugs them.

Meanwhile, this trivial-to-make change would let users make their own
choices about mesh versus longevity; and will greatly extend both live
and suspended battery life for 80,000 non-mesh G1G1 nodes.  (And if
any schools deploy with access points rather than mesh portals, then
those schools win longer battery life too.)

	John