[net2-wg] Meeting Notes Jan 26th

[Philip Levis]

On Thu, 2006-01-26 at 11:04 -0800, Joe Polastre wrote:

> 
> How can you define an interface without a surrounding architecture? 
> That simply does not make sense to me.

There is an architecture, of sorts, but a very limited one: AMSenderC,
AMReceiverC, etc. (TEP 116). But it's quite minimal. What's the
architecture of the protocols in 1.x?

> Besides, how do people put new protocols into TinyOS 2.x if there's no
> architecture?

They build on top of the 2.x packet abstraction.

> And third, why can't the outside community participate in the
> definition of a network architecture?

If my response came across as suggesting that people in the community
can't participate in defining a network architecture, that was not my
intention. My concern was that there was a conflation between Berkeley's
research projects and the agenda of the net2 WG. Of course, there will
be cross-pollination, just as there will be cross-pollination between
the net2 WG and efforts at MIT, USC, EPFL, ETH, Harvard, etc. But the
two (Berkeley research and the net2 WG's efforts) are not the same.

When the net2 WG formed, it had a very limited agenda and charter. I
thought that starting with a grandiose one was problematic: it's better
for the group to start small, get on its feet, and based on those
efforts, decide which directions it wants to go into. For example, once
we're done with collection and dissemination, I could imagine the net2
WG amending its charter to move in one or more of several directions:

1) a network architecture
2) ZigBee
3) a send pool interface

But I think the direction the group takes is up to its members and what
we're interested in pursuing. The reason why collection and
dissemination are the starting charter are because they're reasonably
well understood protocols that are critical to sensornets, and therefore
are a good way to get the ball rolling. The discussions over them have
let us reach a consensus and start working from the same page.

For example, the decision to go for full AM virtualization with fair
queueing was pretty much made by the net2 WG. We shelved the send pool
question because it seemed like a very thorny one, and getting some
experience working together before tackling it appeared to be a better
course of action.

Phil