[Tinyos-devel] [Tinyos Core WG] TEP105: LowPowerListening DutyCycles

[Andreas Koepke]

Philip Levis:
> 
> On May 19, 2008, at 10:43 AM, Andreas Koepke wrote:
> 
>> Philip Levis:
>>> On May 19, 2008, at 10:12 AM, Andreas Koepke wrote:
>>>> Should we go ahead and define such an interface?
>>> The issue here is distinguishing failure cases. I'm leery of defining 
>>> and interface where FAIL means "not acked," because it does not let 
>>> the higher layer distinguish "I am broken" from "sure, retry." E.g., 
>>> CTP would respond very different, with respect to retries and its 
>>> link estimation, between the lower stack failing and a packet not 
>>> being acked.
>>> Phil
>>
>> Currently the only reason to fail is: "Could not access the medium in 
>> time", right?
> 
> No. What if the radio does not respond properly to commands? E.g., the 
> voltage is too low?
> 
> Phil

Ok -- so the sendDone is really an indication. So where should we put 
the confirmation? One solution would be an AMSend (with the indication 
part) and a ReliableAMSend for the confirmation part, so two sendDones 
from two different interfaces for each packet. That seems ugly.