Re: [Roll] ETX metric

[Henning Rogge <hrogge at googlemail.com>]

Am Donnerstag 15 Oktober 2009 22:17:36 schrieb Alexandru Petrescu:
> Henning Rogge a écrit :
> > Am Donnerstag 15 Oktober 2009 21:44:44 schrieb Alexandru Petrescu:
> >> How about software implementing IETF standards, geared towards advancing
> >> on the Standards Track?
> >
> > The RFC conform implementation of OLSR does not work well. The
> > Freifunk/Funkfeuer community networks (which are some of the largest
> > deployed mesh networks) all use ETX because it does not work with
> > hopcount.
> 
> Well it seems like a large effort indeed (you claim it as some of the
> largest deployed mesh networks) - can I ping it?
Most freifunk/funkfeuer networks are hidden between NAT routers that connect 
the networks to a DSL line.

I think Berlin was up to 700+ nodes some time ago, but it's a little bit 
shrinking because DSL is easier to get today.

Leipzig is and Vienna have hundreds of nodes too (Vienna 400-500 I think).

> >> What are we talking about here?  I more and more feel we live in very
> >> different worlds.
> >
> > OLSRv2 WILL contain integrated metric support because everyone in the
> > MANET-WG agrees that it's necessary. We are discussing about if we
> > integrate ETX as the easiest (and layer 1/2 independant) metric as an
> > appendix/example into the main draft or not at the moment.
> 
> Well that sounds as good news for ETX and OLSR.
> 
> Do you have some comment on link energy metrics?
I think both link energy metrics and ETT have the potential to be a much 
better metric than ETX. OLSR.org has ETX at the moment because it's the only 
metric we can support on different OS/hardware platforms. We hope to move to a 
generic "linklayer data framework" soon, so we can start to experiment with 
ETT or similar metrics.

Link energy metrics for WLAN are a little bit difficult because there are no 
exact tables how many energy each specific hardware consumes for a package.

Henning Rogge

Attachment: signature.asc
Description: This is a digitally signed message part.