[net2-wg] Should CTP header have hop count?

[Rodrigo Fonseca]

I will propose a vote: people should vote +1, -1, or 0 (don't care).

In the forwarding path, THL tells us about how far a packet has
traveled, so there's no need for a change there.

So here's the question:

Do you think we should add a byte to the route control packets that
includes the hopcount, using the current parent, to the closest root?
This is in addition to the path ETX metric already in the packet.

  This requires adding a field to the route control packets, 1 byte
should be enough.
  The code complexity to do this is minor.
  It can be exposed through a RouteControl interface. Other protocols,
like BVR and PathDCS use the hopcount for localization. I've seen
hopcount used for neighbor selection (as in you need to have at least
one neighbor that is further from the root than you), but pathETX
could serve the same purpose.

  My vote is +1, as it may be useful, and the cost is not high.

Others?

Rodrigo





On Mon, Nov 17, 2008 at 2:36 PM, Omprakash Gnawali <gnawali at usc.edu> wrote:
> On Fri, Nov 14, 2008 at 10:23 AM, Omprakash Gnawali <gnawali at usc.edu> wrote:
>> On Thu, Nov 13, 2008 at 7:13 AM, Rodrigo Fonseca
>> <rfonseca at cs.berkeley.edu> wrote:
>>>>
>>>> I don't think there is a way for the origin to estimate its distance
>>>> from the root for this experiment.
>>>>
>>>
>>> Ah, so you are talking about adding hopcount to the control packets?
>>>
>>> In that case, you are right, we don't know the distance from the root.
>>> To know that, we would have to break the ETX field in two:
>>>   * a) hopcount (each node in the tree adds one)
>>>   * b) extra ETX (cumulative extra ETX per hop)
>>>  such that a + b equals the current ETX.
>>>
>>> For the data path, I don't think hopcount is necessary, or different
>>> in practice from THL.
>>>
>>> Rodrigo
>>>
>>
>> Exactly. Prior protocols (including mlqi in TinyOS 2.x) exchange
>> separate path metric and hopcount in control packets. They also export
>> the hopcount via Routecontrol. It would be good to know if people had
>> some use cases in mind for that.
>>
>
> Alright, my sense is we are not thrilled about hop count.
>
> - om_p
>