[net2-wg] TEP 118

[Philip Levis]

On Feb 8, 2006, at 1:47 PM, henri dubois-ferriere wrote:

> Looks good to me. A few small comments.
>
> 1) If there must be one, I would suggest that the tie-breaking
> mechanism be specified by the TEP.
> If I write an app where ties may happen, I would like to know what is
> the algo to resolve them. [ i know, this raises the question "are we
> defining interfaces or protocols"...;) ]
> Now, on thinking it over, i don't see an algorithm for tie resolution
> which is guaranteed to converge. I could be tired, or stupid, but can
> anyone formulate such a thing?

Ties are broken by the nodeID of the originate node. Assuming nodes  
have unique IDs, then the largest ID will always win. Essentially the  
version# is a concatenation of the version and the nodeID.

I think that this gets to the crux of the question of whether we are  
specifying protocols or interfaces.


> 2) could the TEP describe how one would go about disseminating with
> multiple sinks/basestations? This could be a good
> illustration/validation of the interfaces.

Yes, good point.

>
> 3) are these interfaces also intended for "all-to-all" dissemination,
> where any node in the network might have stuff to disseminate? it
> would be nice to make this explicit.

It is. OK.

I'll try to have an updated draft by the next telecon.

Phil