[SensorNetArch] Question: How do we fit timesync protocols over SP?

[David Culler]

Yep, I think there is a list of "issues yet to resove" where we can 
articulate a design viewpoint but recognize that we haven't fully 
resolve it realization.

Joe Polastre wrote:

> I decided (with Prabal) that rather than "half-ass" what the SP
> interface is for timestamping in the next 2 days, instead we mention
> it in the "design limitations" section of the "implications" section
> with a proposed solution.
> 
> -Joe
> 
> On Apr 6, 2005 10:51 PM, Jerry Zhao <zhao at icsi.berkeley.edu> wrote:
> 
>>Yes, I totally agree timestamping should be specified in SP layer,
>>nowhere else to go. Since it is closely tied with underlying MAC/MAC-
>>adapter, it can just be mandatory in the interface.
>>
>>On Wed, 2005-04-06 at 22:04 -0700, prabal at eecs.berkeley.edu wrote:
>>
>>>The Case for Timestamping Support in SP
>>>=======================================
>>>
>>>- Timestamping is required for some middleware services / applications
>>>- Timestamping in MAC layer is necessary for low jitter / high precision
>>>- Timestamping information crosses the narrow-waist
>>>- Timestamping in the MAC/link layers is often available and nearly free
>>>- Timestamping should be included in the SP abstraction
>>>
>>>- If an application or middleware service requires time synchronization or timestamping, then a MAC which does not support timestamping (or can only support it in a very coarse-grained manner) will be a poor choice for the application and is unlikely to be chosen for the application.
>>>
>>>- Timestamping may or may not be supported by underlying hardware/MAC
>>>- Timestamping quality may vary based on the degree of hardware/MAC support
>>>- Timestamping can be approximated if MAC doesn't support it directly
>>>- Timestamping quality, even when approximated, should have bounded error [T?]
>>>- Timestamping can be supported by the link abstraction w/o a loss of generality
>>>- If SP is not the right place to abstract and expose timestamping, what is?
>>>
>>>Prabal
>>
>>
> _______________________________________________
> SensorNetArch mailing list
> SensorNetArch at Millennium.Berkeley.EDU
> http://Mail.Millennium.Berkeley.EDU/mailman/listinfo/sensornetarch