Re: [Roll] ETX metric

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

On Oct 15, 2009, at 6:22 AM, JP Vasseur wrote:

> On Oct 14, 2009, at 9:49 PM, Philip Levis wrote:
>
> > I think we should keep in mind that RPL must work across multiple  
> > link layers. E.g.:
> >
> > A -- 900MHz --> B -- 2.4GHz --> C
> >
> > ETX is a commonly used metric, but typically under two simplifying  
> > assumptions:
> >
> > 1) Routes are composed of a single link layer
> > 2) The link layer has a static bitrate
> >
> > If you break either assumption, it doesn't work well. ETT (expected  
> > time of transmission) can address 2), but not 1).
>
> Not sure I would agree on this statement though. Being based on  
> probing it's not link layer dependent.

You can of course use ETX in a link-independent way. It just leads to  
possibly very bizarre route choices. E.g., choose a very slow, high  
power link that requires one transmission over a very fast, low-power  
link that requires 2 transmissions. ETT breaks when you have a fast  
very high power link and a slower low power link. Nominally, one goal  
of ETT and ETX is to minimize the space-time that packets occupy a  
wireless channel. When different link layers have very different  
transmit ranges (space), they stop being accurate units for this  
purpose.

Did I understand you correctly?

Phil