Re: [Roll] ETX metric

[Philip Levis <pal at cs.stanford.edu>]

On Oct 13, 2009, at 9:02 AM, Richard Kelsey wrote:

>   Those who cannot learn from history are doomed to repeat it.
>
> We have to be careful not to learn the wrong lesson.
> Hopcount without careful link assessment definitely doesn't
> work.  Hopcount limited to stable, symmetric links is a
> different beast entirely.  For example, consider the Cost/PL
> column in table 1 of [1], which gives the average
> transmissions per link.  For most of the networks it is very
> close to 1 and it never goes above 2.  That's puts ETX
> pretty close to hopcount for those networks, if routing is
> limited to links with good ETX.

Richard,

I fear this assertion might misrepresent the data. An average cost of  
<2 does not by any means suggest that there are no links with cost >2.  
It could be that 80% of links have a cost of 1 and 20% have a cost of 4.

Phil