[net2-wg] Meeting Notes Jan 26th

[Jonathan Hui]

Om: Collection people have been emailing each other. Have a plan in place.

Rodrigo: Which collection layer should we start from? Mulihop? MultihopLQI?

Joe: Suggest writing collection layer from scratch. The interfaces
need to be redone. The forwarding engine also needs to be redone. One
example is that multiple parents is not very well supported.

Rodrigo: Thinks that maintaining ETX to the root is good.

Joe: We've tried ETX and it has some reliability problems. Really
suggest having multiple parents to improve reliability.

Rodrigo: There's a getNextHops command.

Joe: Yeah, don't want to wait for you parent to go away before looking
for another one. Have to deal with loops when choosing the next
guy. MintRoute and MultiHopLQI punt on the loop part.

Om: Has anyone notified Kristin Wright? Someone should notify her if
we haven't.

Joe: Skeptical that this stuff is going to be done by next Wednesday.

Om: Should we be testing our code on mica2's as well?

Gil: We've been mainly focusing on the micaZ and haven't done much
mica2 work.

Om: mica2's would be a problem.

Rodrigo: Still need to talk about link estmation. Ideally would like
to separate it from the network layer.

Joe: Hopefully link estimation is no longer being done in
MintRoute. Sound like there will be no SP in this release either.

Joe: Beacon-based is always recommended. The best is a hybrid
approach. Want to use RSSI/LQI to bootstrap the link estimates and
then use a beacon based approach to maintain the link estimates.

Rodrigo: For TTX, we need a story on ZigBee.

Gil: Anyone who has the passion for ZigBee can certainly contribute a
ZigBee layer.

Joe: I thought this group was about a network architecture.

Om: We had a discussion on architecture, but we needed to get
something out sooner. We had the whole discussion on SP, message pool,
etc.

Joe: It seems like if tos2 is to be adopted, most of the innovation is
going to happen at the network layer. So a good architecture is really
crucial. I think Phil is right in that we need 2 example protocols,
but the real value is how to start composing the protocols.

Joe: Sounds like we need to move forward and try and get something
out. I would be glad to help with refining interfaces and stuff like
that.

Rodrigo: Important that anything we do does not preclude ZigBee.

Joe: Should be able to answer the ZigBee question if it comes up.

Gil: I can follow up with Phil on the dissemination stuff.

Rodrigo: Can follow up with Phil on collection stuff.

--
Jonathan W. Hui
jwhui at cs.berkeley.edu
http://www.cs.berkeley.edu/~jwhui/