Net2WG 20070222 Meeting Note
*net2wg Future Directions
Om (O), Phil (P), Rodrigo (R), Sukun (S) (alphabetical), Note: Phil, Sukun
O: If everyone could send me when they are available, then I can schedule a new time so Arsalan and Kyle and others can meet.
O: First item is net2 future directions. There wasn't a whole lot of responses.
R: There was ZigBee.
O: ZigBee popped up again. And Phil sent an email.
P: You proposed pulling people in as a way to get their protocols into the three.
O: Yes, that could be a great way to recruit people for the core protocols that we maintain.
R: So what is the set of protocols that we maintain? What should be the set of protocols in the networking stack? Collection, dissemination, what else?
O: How about a ZigBee stack?
R: What about 6lowpan?
P: That sounds much more interesting to me.
R: From the perspective of IP, it could be an overlay over some sensornet routing protocol. You give an IP address, and it looks like it's one hop. But the mesh header implies that it needs some kind of mesh routing algorithm, point-to-point, of which there is no specification.
P: Each node has an IP address, unique ID (e.g. 16 bit ID). There are a couple of specified head format for fragmentation and assembly, mesh routing. One thing that can be specified is segmentation and assembly is done at end hosts.
R: Is it from node to PC?
P: End-to-end in the same network.
P: Mesh routing header come before fragmentation header. There is an idea of sending IP packet single hop.
P: We can do CTP on lowpan6 CTP would work as an overlay over IPv6. I am not sure whether IP would be happy with that.
R: With mesh header, there is some mesh routing algorithm. There is no specification, right?
O: I don’t think we have ever really done in TTX. Is there some portal or link?
R: It could be some message that net2wg has done lots of work and concluded on dissemination and collection. Now we are moving on, so what it should be?
P: If we get into the world where we do what others want to do, not what we want to do…
R: That’s what I get, not the former.
O: As the researchers, what is the network research?
P: Good next step for CTP is low power. It gets back to what is the interface, preamble sequences, scheduling of AM sequence, etc. in low power.
P: Boomerang has some way in which what would be in neighbor table. TDMA implies fundamental challenge, like the size of neighbor table.
O: If we decide to probe, what would be the place?
O: We can work on a draft and send out.
R: How is the course in net2?
P: One thing SP is appealing is that you can do your own scheduling.
(Talks about grant to send)
R: Standardizing link estimator, then I can build BVR without using its own link estimator.
O: I talked to IFRC, they say it is already a solved problem.
P: Like BVR, and there are lots of protocols out there.
R: The other thing about dissemination. There are three types: really short (1 packet), order of 10’s, and really big one like a program.
O: We just address short one.
R: There is a command dissemination.
O: We can send a draft out to net2wg. We can send out a note and ask what people think.
O: I wish to talk pending issues. One was inter-packet time. And there is queue information.
R: We can do Flush on top of it.
O: Providing interface to do that.
P: Rather than support on the queue information, it would be also helpful some event. What if queue is empty?
O: A way to get current working group more involved: send an interface TEP, and say ‘please comment’.
P: Collection interface? Or SP interface?
O: Collection and dissemination.
O: Many people might be already part of other working groups.
O: At least specify interface for inter-packet time and queue information.
R, S: We will do by next meeting.
P: It will be great to have BVR on T2.
R: We have a location service working on BVR.
O: What kind of location service is it?
R: BVR paper actually describes it. (Note taker was not following clearly… using hash and beacon coordinates…)
O: I am interested in low power dissemination and collection.
S: I am also interested in low power collection, and finishing Flush which is pending for a while.
R: Point-to-point and Control.
O: Parking application would require point-to-point. Let’s assume we want to solve this problem in a distributed manner, we need a point-to-point routing.
P: Sounds like perfect for geographic routing.
O: Okay we have some ideas.