[Tinyos-devel] timestamps in seriallisten/sflisten etc.

[Vlado Handziski]

On Thu, Jun 26, 2008 at 22:32, Philip Levis <pal at cs.stanford.edu> wrote:

>
> On Jun 26, 2008, at 1:29 PM, Philip Levis wrote:
>
>
>> On Jun 26, 2008, at 1:21 PM, Vlado Handziski wrote:
>>
>>  Yes, there is currently no proper hanshake apart from the
>>> possibility of selecting an ACKed or not ACKed packet. I don't think
>>> we have a rule about this, it is just a convention that one
>>> direction was ACKed and the other was not. But by using only 2
>>> symmetric ACKed packet exchanges (one mote->PC, the other PC->mote)
>>> we can get a very good estimate of the propagation time.
>>>
>>
>> TEP 113 goes into why ACKs only occur in one direction. The assumption
>> is that packet losses are mostly due to overflows and not bit errors.
>> So PC->mote has a much higher chance of loss than PC->mote
>>
>
> er, should read:
>
> "So PC->mote has a much higher chance of loss than mote->PC..."
>
>
I agree that most of the time we don't need the ACKs in the mote->PC
direction, so the default can remain as is. But we definitely should enable
the possibility to request an ACK in the other direction. We just need a
single retransmit timer. I don't think it is a significant bloat.

Vlado
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://www.millennium.berkeley.edu/pipermail/tinyos-devel/attachments/20080626/718fd059/attachment.htm