[SensorNetArch] Supporting schedules?

[Cheng Tien Ee]

Hmm, from Barbara's paper it looks like FPS was built with the 
assumption that it runs over CSMA. It didn't require fine-grain 
synchronization, so the basic interface was fine, although she did 
mention in passing that if fine-grain sync is required some sort of 
timestamping service from the MAC is required. I can talk to her about 
it and email what I get.

>Ok, from my perspective, we just spent most of a meeting talking about
>interfaces for scheduled communication. Is the goal of SP to be a
>potentially-scheduled link-layer interface, or to be a link layer
>interface that can _support_ scheduling of communication by a separate
>layer?
>  
>
To me, it should be the latter.

To tell the truth, I've always had the impression that doing real-time 
over wireless is hopeless, primarily because of the low link quality 
(compared to wired networks). But I would agree that if the lower layer 
cannot provide a certain level of service, then the higher layer will 
most likely not be able to do it as well.

One of the tests to determine whether or not the architecture is 
successful is its adoption by the community. I can imagine people who 
look at the interface and wonder what sort of values to put in. Perhaps 
we can have different levels of information passed by the interface? So 
if I'm interested in best effort traffic in the network, then I'll just 
use the plain old sendmsg / recvmsg interfaces. I think Phil covered 
this in his write-up(?).

>Gil, just trying to break the deadlock.
>  
>
Thanks...

Ee