Net2WG/Notes/20060629
Present: Arsalan (A), Geoffrey (G), Phil (P), Rodrigo (R), Sukun (S) Note: Sukun
Loop handling, link estimation, neighbor table management, and release were discussed.
P: There was a discussion in T2, for release: Should AM packet have source address? Eventual conclusion was that it should. So release version will have get_source call.
P: Kyle, Om, Rodrigo and I have been banging on collection, it is pretty much there.
R: We’ve been testing link estimator parameters. Yesterday, we had problems. In the morning, 88~90% of packets were coming. 45% in the afternoon. After 6:30pm, it went back to 95%.
P: Which channel?
R: 26. Intel has a team testing 802.11. They may have set the power too high.
P: Even at channel 26, there can be interference.
- There was a discussion about loop handling, which I could not follow well. Phil and Rodrigo can talk more.
R: We would need to get more tests in different environments. Losses were correlated. We still see a lot of loops.
P: When the network condition is good, 90%~95% end-to-end success rate. Failure comes from loops.
R: Every time a node sends a packet, it put gradient.
P: It is good to have information in packet not in node. Keeping track of all packets in the node is not good. When a loop is detected, we can stop forwarding for a minute.
P: Routing engine is not controlling link table. It can only pin the parent.
A: In SP, you can pick which one to remove.
P: What happens if another protocol wants that node in the table?
P: Rodrigo says worst link out. Arsalan says putting policy on the top.
P: Arsalan is right: why link quality isn’t the best (link quality is a good metric).
R: It is the only metric which can be inside the link estimator.
R: Is RPM out?
P: Yes. It is beta2-candidate. We will merge it with changes made after making the RPM. We test everything in RPM, make another RPM, and iterate over RPMs.
R: tinyos.jar has not been updated.
- There was a talk about descendants between Phil and Rodrigo, which I could not follow. Sorry.
R: We are getting 95% of packets, so it would be okay to release at beta release. How could I test RPM?
P: I will let you know.
R: I was testing packet injector. After sending a packet, I could not hear packets.
P: We haven’t tested dissemination in parallel with collection.
R: What pattern does dissemination have? Trickle?
P: Yes.
R: It would be great to see both (collection and dissemination running).
P: We need to sit down, and write down protocols. One-to-one correspondence between code and protocol will be good.