[net2-wg] Minute: Teleconference 1/19/06

[Rodrigo Fonseca]

TinyOS 2 Networking Group
Meeting 20060119
Scribe: Rodrigo

Collection Signature
========================
Phil: any updates on the collection signature?
Om: use the standard send and receive.
Phil: For the root of collection tree, receive will be signalled?
How do you know if you are the root of a collection tree?

Om: might be dynamic or not. I looked at the interface at t2.
Phil: Interfaces on TEP 116. T2 interfaces were made to agree with the
TEP, can be assumed to be that.

Tweaks: send and receive have accessor methods for payload and

Kyle: cancel is present in the new intefaces.
What does it mean to cancel a message on its way to the root?
Phil: You don't need to implement cancel. Cancel semantics: if you
can't deal with it, return FAIL. cc2420, after the packet is
submitted, you can't cancel.
Om: there are only two changes.
Phil: yes, only added these accessors, so that when you send you don't
have to wire all interfaces.

Om: will there be test applications?
Phil: there should be. RadioCntToLeds deals with the new interfaces.

Polastre: protocol over the UART, or some external force, to tell a
node it is the root of a tree.

Phil: you are the root, should be compile time or at runtime.
Polastre: exposed as an attribute, getAttribute
Rodrigo: Where is that?
Polastre: tep114

Rodrigo: if that is an interface, one of the mechanisms to activate it
is if you are connected, but there can be other ways of calling it.
Gil: Is that an event or command?
Polastre: can be both
Gil: decouple that a little bit

Rodrigo: the bottom line is that it should be possible to become the
root at runtime, and possibly cease being a root.
Phil: should there be a compile time as well?
Om: makefile

Rodrigo: A very simple component could call the interface at the init time
Phil: what is the default case?
Om:on the testbed all motes are connected to the root

Gil: TOSBase as the root of the tree? As in Drain?
Om: the root of the is just a relay.
Polastre: nothing prevents that. TOSBase can just emulate what the root does.
Phil: just define the protocol
Om: is that a problem if you have only motes?

Phil: specify a protocol, and a PC side tool can perform it.

Next item: Dissemination and Collection Progress
================================================
Phil: checked in code to implement Trickle timers
Collection: still tweaking side cases.
Gil: skeleton code that uses the interfaces.
Phil: should be working today

ZigBee
===========

Joe: email to T2 Group. It would be good to have a ZigBee
implementation. Lend credibility. The network arch will have to run
ZigBee. Get a document together: how would Zigbee work. XBow is
interested, Moteiv is interested.

Gil: seems to require implementing the 15.4 MAC
Joe: it's simple.

Kyle: summarize
Joe:* subset of 15.4 implemented.
    * Coppenhagen: interfaces done
    * Good defined link interface
    * ServiceS: networking service, collection and AODV. TinyAODV
    * Security services

Rodrigo: documents?
Joe: zigbee.org
Phil: download and read freely, for non-commercial purposes
Joe: non-comm. academic purposes. Companies have to join the Alliance,
but academia is free.
Joe: can benchmark and validate what we are doing.

Om: People who want to do research on MintRoute, how can they benefit?
Joe: People can benchmark against the implementation. recommendations
back to Zigbee. Opportunity to participate. TOS Alliance should become
a member of the ZigBee alliance.

Kyle: is Zigbee tied to 15.4?
Joe: we can do a decoupling from 15.4. You could run it on top of
other radios, using possibly SP.

Gil: icky design decisions. Whoever's implementing would have to ignore those?
Joe: possible to restructure it. Would benefit everyone.

Kyle: overlap with the network stack. Is this to be the defacto
networking stack.
Joe: show that ZigBee can be built on the T2 architecture.
Joe: has useful protocols. Components of Zigbee that can be used in
other contexts if we have the right modularity. Specially the security
services.

Phil: useful modifications to the link layer from security
Joe: three layers. Can choose which ones to do. Forces us to think
about security as it relates to SP.

Phil: whether there are people who want to push through. Coppenhagen
should be involved perhaps. The set of people interested in TinyOS
protocols and ZigBee may have non-trivial overlaps.
Joe: XBow is very interested.

Phil: we should go read the spec.
Joe: I'll send pointers

Rodrigo: it will show that the architecture is good if accomodates
ZigBee, and we shold at least aware of if when designing it.

The end.