[Tinyos-devel] retransmissions in networking protocols

[Philip Levis]

On Oct 5, 2007, at 2:00 AM, Andreas Koepke wrote:

> Just one more remark on where to put the L2 retransmissions (MAC vs.
> routing protocol).
>
> It seems that we have a funny situation: because net and core did not
> yet agree on how to communicate link costs from PHY to anyone, all net
> protocols in T2 usually assume a CC2420 radio, often with the  
> additional
> assumption that it is a plain CSMA and not a preamble sampling CSMA or
> some other energy saving MAC protocol is used. The first assumption is
> easily to be seen in the wiring, the second one in the way how
> retransmissions are done (lqi retries immediately, ctp tries to be  
> a bit
> more platform independent, but retries after 4-32ms).

Please look at the CTP header file -- the retry timing depends on the  
platform.  Admittedly, it does not consider all possible platforms.  
Having some general mechanism to learn a "packet time" from the link  
layer would help here.

You're right that CTP currently does not necessarily work well once  
you add significant latencies (e.g., low power); now that we've  
figured out the link estimation and route selection part, packet  
timing and low-power is the next item on our agenda.

Phil