Present: Omprakash Gnawali, Kyle Jamieson, Sukun Kim, Mike Liang, Razvan Musaloiu-E.
Om: Any reactions to the net2 work? Kyle briefly went over the program and mention the question about reliability raise by someone from Sensicast.
Om: How was the call for participation received? Kyle: Hard to say because the audience was not very responsive. Some people push me afterward and enthusiastically ask about networking protocols. I think there is already a flow of people coming in. The were no objections raised (but the audience was pretty sleepy).
Om: Did the core said anything relevant to net2? Kyle: A couple references but nothings that impact us.
Om: Are the slides going to be up? Kyle: I'm pretty sure they will be. I gave David Culler my slides.
Kyle: There was also a contrib session. Om: Any network related things? Kyle: I can't remember. I think they were. To be honest, I don't remember. But the slides should be up. Kyle: The people that might be interesting to work with are the Koreans. KETI institute. They are very enthusiastic and they are doing a lot of stuff. Om: Did they do any presentation during the contrib? Kyle: They made a presentation. They were on the panel. I think they talk twice: once in contrib and one on the panel. Om: What did they talk about? Kyle: I think they do network related stuff. Om: I would be nice to know what they do... perhaps they are doing something unique... Any other impressions? Sukun? Sukun: It was a more in depth version of last year. They were more technical items.
Om: Let's talk about Deluge. I'm glad to hear the T2-dissemination is used. Mike: I will start with a brief overview of how it works. You have a base station connected to PC, which you first push program images to. Then, the base station is instructed to start network-wide reprogramming, and it uses T2-dissemination to notify the network of the availability of new images. Upon receiving this T2-dissemination message, motes start Deluge to grab program images over-the-air and reprogram themselves.
Om: How is this different from the original, T1, Deluge? Mike: We try to keep everything similar to T1 Deluge from the user's perspective, such as outputs from the tools. However, there are some differences in the architecture. One thing is that we try to make a stronger separation between the reprogramming component and network service component. This allows users to replace the network service component, or use it in their applications if they like. Other differences include python tools, instead of java tools. Razvan: Right now, the interface between Deluge and storage is based on BlockRead/Write. If somebody wants to use Deluge to disseminate, he can just implement BlockRead/Write. Om: I see. So, if somebody wants to disseminate 50 bytes of data, he can just write a module that uses BlockRead/Write. Razvan: One limitation so far is that Deluge can disseminate only one image at any given time. We couldn't find a clean way to multiplex Deluge.
Kyle: If a new program image breaks the radio, then you can't reprogram remotely, right? Razvan: If you have physical access, you can load golden image by pressing user button in a certain way. Om: If you set up a mechanism in which you send messages periodically, and if that doesn't happen frequently enough, you automatically reboot the mote. Mike: That's possible.
Om: What kind of testing have you done so far? Mike: We have done small-scale testing, and we will test it in our testbed. Om: Do you have any design document or something like that? Mike: We don't have such a documentation yet. However, we will do this after testing.