[Tinyos-devel] Re: [Tinyos Core WG] time sync interfaces

[Janos Sallai]

>> In 90% of the cases, a sensor network is homogeneous: all devices are
>> identical and programmed with the same program image. That is, the
>> issue
>> of interoperability of different timestamping implementations
>> simply does
>> not exist.
>
> There's a difference between an artifact of deployment approaches
> today and application requirements. The former will change
> significantly as networks grow in complexity, longevity, and scale;
> the latter won't.

I completely agree with that.

>> Prescribing a common "network time precision" would put extra
>> burden on
>> the implementer of a radio stack, and could require "extensive"
>> computation in time-critical code.
>
> At some point, you have to make a call. Of course, if the standard
> expected time precision cause extensive computation, an implementer
> might decide to not support it. It's just that having at least one
> statement of a standard (but not necessarily omnipresent) timing
> fidelity allows systems to extend and build on prior work, rather
> than the unmaintainable state of having to write everything yourself
> for every application.

Not necessarily. The precision of the timestamp in the footer can be
arbitrary (i.e. platform specific or radio stack specific), as long as
there is a way for the recipients of the packet (which are potentially
different devices) to figure that out. This would encourage implementers
of radio stacks not to omit timestamping support just because complying
with a "network time precision" is too cumbersome.

The precision (and width) provided by the HIL is completely independent
from settling on an over-the-air timestamp precision. Well, completely
independent might be an overstatement, but for sure, they don't have to be
identical. Of course, the over-the-air precision is equal or higher than
the HIL precision required by the standard. We can well say that the radio
stacks MUST provide PacketTimeStamp and PacketTimeSync with a prescribed
"standard precision", without specifying a standard over-the-air network
time precision.

It's trivial to implement a component that converts the timestamps from
radio stack specific over-the-air precision to "standard precision". If
the network is homogeneous, you don't even have to ask the sender what the
timestamp precision of the received packet is. (This is what I had been
referring to when I was writing that 90% of the deployments are
homogeneous.)

I expect that timesync protocol implementations in T2 (e.g. FTSP, RBS,
RITS) wouldn't wire the "standard precision" interfaces, but would work
directly with the arbitrary precision ones the radio stacks provide. I
would say that in 90% of the cases applications would wire the above
protocols instead of the low-level timestamping primitives. For the
remaining 10%, I don't think it's worth specifying a standard precision
for the HIL, either.

Janos

Janos
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3096 bytes
Desc: not available
Url : https://www.millennium.berkeley.edu/pipermail/tinyos-devel/attachments/20071219/772de97e/smime.bin