(CCing the OLSR list, since George also asked the same question there but got no answer yet)

On Dec 9, 2009, at 9:42 PM, George Sanders wrote:

I am researching meshes/MANETs, and the accompanying 802.11s/HWMP/olsr/etc.

Before asking any specific questions, I wonder if someone could point me to some proper OSI diagrams that I cannot find - specifically:

ok, here is my attempt.....

+--------------+ | Application | 7 +--------------+ | presentation | 6 E_DOES_NOT_REALLY_EXIST +--------------+ | session | 5 E_DOES_NOT_REALLY_EXIST +--------------+ | transport | 4 +--------------+-----------------------------+ | IP, OLSR.org, B.A.T.M.A.N layer 3 version | 3 OLSR.org's OLSR implemenation is located here +-------------------------------------------------------------+ | link, 802.11s, B.A.T.M.A.N layer 2 version, OLPC's 802.11s| 2 the HWMP and all the 80.211s stuff resides here. 802.11a/b/g/n resides here. Madwifi, ath5k and ath9k reside here. There is massive room for improvement on the Wi-Fi level! +--------------------------------------------+----------------+ | phy | 1 this part of the stack is definitely a problem for any routing protocol [*] +--------------------------------------------+

[*] Much of the discussion focuses on "protocol $bla is better than $foo" but actually the Wi-Fi layer 2 and 1 could need massive improvements! It is not the routing protocol on higher levels which matters. Wi-Fi per se was not intended to scale so high as we use it in community Wi-Fi networks :) In case layer 1 and 2 were very good and stable, we would be able to run almost any protocol on top .

• can someone draw, or direct me to, an OSI diagram showing an 802.11a/b/g/n network using

plain old vanilla 802.11s + HWMP (no OLSR) to support an IP network ?

Then, for comparison:

• can someone draw, or direct me to, an OSI diagram showing an 802.11a/b/g/n network using

ONLY BATMAN (no 802.11s) ... ?

Have to ask back: which version do you mean?

Please see: http://www.open-mesh.net/wiki/BranchesExplained

Finally, in my research I came across this very interesting conversation from 2006:

http://www.ietf.org/mail-archive/web/manet/current/msg08218.html

and I wonder, has that question:

"I agree, but we may still be able to pull it off without having to messup the stack."

been answered yet ? Or is there still argument over whether or not some implementations require "discussion" between components on different OSI layers ?

In my personal opinion, the more we can get *standardized* "hints" from the lower layers (especially Wi-Fi), the better any routing metric can become. In OLSR.org the metric calculations were offloaded to a plugin architecture so you are quite free in this respect. I guess there is something similar in some versions of B.A.T.M.A.N. and I am sure people on the BATMAN list can actually answer this question better than me.

Thank you.

I hope it helped somewhat.

a.

PS: typing B.A.T.M.A.N. with all the "." dots all the time is... tiring. How about renaming it to "Span"? What came out of that discussion?