[Commotion-dev] Throughput advantage to 802.11s support

[L. Aaron Kaplan]

On Feb 8, 2013, at 5:42 PM, Ben West <ben at gowasabi.net> wrote:

> Hi Dan,
> 
> First off, I believe all of these patches were made against the current version of the compat-wireless library.  The first patch did happen to apply cleanly to compat-wireless-2012-09-07 integrated into Attitude Adjustment, but the second patch failed.
> 
> This first patch is to add short slot time support to 802.11a adhoc mode (i.e. 5GHz only).  Long slots are required for 802.11b/g adhoc mode (i.e. 2.4GHz) to maintain compliant with the standard.
> 
> https://patchwork.kernel.org/patch/1835531/
> 
> This second patch, on the other hand, will allow short slot slots in 802.11s mode in 2.4GHz and 5.8GHz, which the standard allows.
> 
> https://patchwork.kernel.org/patch/2050741/
> 
> I do understand Sascha, et al, have expressed criticism for the fact that the 802.11s draft stipulates an arbitrary maximum of 32 nodes per mesh.  (Thanks to whichever dastardly telco lobbyist...)

That's partly correct today. Initially the node_id field was 5 bit long -> hence allowed for a maximum of 32 nodes.
I was told cozybit changed that in the mean time (are they then standard conforming?)

What I was told behind the scenes: it was not a telco lobbyist, but actually the companies involved in standardising 802.11s themselves, which did not want it to be much of a success since all of them already had their proprietary protocol implementation :(
I have no idea if this was a mere opinion of this IEEE forum participant or if it really reflected the situation at the IEEE at that time. However, it would explain brain dead ideas like a "32 bits are enough".

Do you remember the 802.11s implementation at OLPC? That was terrible...they had to revert to fixed APs.

>  However, Antonio of the Robin Mesh project, who has already incorporated 802.11s into this new Meshroot firmware stack, mentioned how this limitation is actually 50 nodes as implemented (??),

... in any case... is 50 nodes much better than 32 nodes?
Not really. Come one :) 

I'd like to see an 802.11s network with > 1000 nodes!
(Probably will never exist since it is one single collision domain).
To sum it up: I am very doubtful of its scalability. 

For small, local layer 2 meshes it might be OK after it got lots of massaging from cozybit - don't know.
We could test it at the WBM in Denmark.
Try it out and tell us about it :)

a.