[Olsr-dev] Strategy for ETT metric

[Henning Rogge]

On Fri, Apr 4, 2008 at 11:34 PM, sebastian sauer <(spam-protected)> wrote:
> Thu 27 Mar 2008 21:50, Aaron Kaplan wrote:
>  like the previous LQ-dispute on which henning and i had different opinions, this
>  here is to a large degree a philosophical || dicatorial decission (you could also
>  say a matter of personal taste || personal compromise)
So maybe we should implement a "switch" for the testing phase and
hardcode it to a specific strategy later.

>  e.g. we could argue - i'm sure ;) - endlessly if either option #1 or #2 is
>  more fair (less bias)..
yes, maybe... ;)

>  <pedantic>
>  my "horror" vision: some nodes configured with option #1 and some with
>  option #2 within the same network.
>  </pedantic>
>
>
>  > At least right now, we don't know what is better.
>  i doubt we ever will.
Maybe a constant rate of "probe packages" is not the best choice. We
can only calculate the bandwith if one package reach the 1-hop
neighbors. On a link with ETX 3 only one in three packages would do
this... should we send more packages (sounds bad in my oppinion).

Another thing I just heared from a coworker is that bandwith is nearly
always correlated with the received signal strength ? Can someone
check this with some tcpdumps (and radiotap support) ? If this is
true, it would make anything a LOT easier.

Henning


-- 
"Wo kämen wir hin, wenn alle sagten, wo kämem wir hin, und niemand
ginge, um einmal zu schauen, wohin man käme, wenn man ginge." (Kurt
Marti)