Rodrigo, Phil, Arsalan, Sukun, Henri, Om, Kyle, Ning
Arsalan: Put together a set of interfaces and structures. Trying to capture the requirements of different layers. NeighborTable would be basic standard information for single hop. You can insert, pin, unpin a neighbor. There is a link handle which is index into the table rather than using the actual address.
Henri: What will be cleaner due to the handles?
Arsalan: Applications. If the address format changes no need to change the app.
Phil: What is the lifetime of the handle as a value? If the handle gets invalidated if the message is in the pool, what happens?
Phil: App looks up the handle and later decide to send to it, how to handle that because it might be invalid.
Phil: Invalidated signal to up? Unique within a certain time. Handle could be certain bits of index + certain bits of address.
Arsalan: There is a mapping of lower layer address and interface type like in Boomerang. For instance, allows to send through the serial interface (UART) rather than send to the UART.
Arsalan: Should the application be asked for an opinion before an entry is evicted? Can we assume a resource manager?
Phil: How are the turn on/off fields useful?
Arsalan: Right now only the next on/off time.
Phil: The original motivation was FPS which did network level scheduling. Still just next time it is on - useful to know and to set when a link will be available. This happens when the MAC as well as the network layer are trying to schedule.
Arsalan: Link estimation, similar to the current interfaces.
Phil: Should it be ETX or PRR?
Henri: ETX seems narrow, for example in CC1100, one can specify encoding which changes the error rate.
Phil: ETX is abstract in the sense that it tells you the cost you have to incur to send a packet across.
Henri: In dutycycling, uncertainties, have to make more effort so need more general metric. Getting back to ETX, the metric should be addititive which makes it easy to compare, compose.
Phil: If two different implementations, then there needs to be an agreement in what the number means. We might calculate it differently, we want them to interoperate.
Phil: Lets say it is ETX. Then if you are using redundant encoding and turns out if you are sending packets that are about twice as large, then you would estimate the ETX to be 2.
Om: How would you know the ETX if you are using LQI?
Phil: Use a function to convert from LQI to ETX.
Arsalan: Message pool. There is an urgent bit. How does it work if there are multiple protocols.
Phil: In Boomerang it is never used.
Arsalan: Another possibility is specifying the deadline for sending. Send after a certain time. Cancel after a certain time.
Phil: Are the timeouts going to be small or big? Is it going to be in terms of packet time?
Om: Real time only if there are real time constraints. Protocols can use just packet time.
Arsalan: Rawsend to send without handle.
Phil: Two different collection protocols. Directories will be created today. It will be good if other people can look at at TEP 123 and 124.
Rodrigo: We will continue discussion on link layer on the wiki and emails.