[Commotion-dev] Roaming between nodes

[Dan Staples]

On 07/04/2013 04:46 AM, Mikael "MMN-o" Nordfeldth wrote:
> On 2013-07-03 16:07, Dan Staples wrote:> I assume you mean
> roaming/handoff of non-mesh clients connected to
>> access points on mesh nodes, yes? Unfortunately, this is currently not
>> possible, without the type of coordination you describe.
> 
> Ah, yes. Handoff/-over is what I mean.
> 
>> As I understand
>> it, it is a limitation of a layer-3 mesh, such as OLSRd, which we use.
>> I've heard that a layer-2 mesh, such as Batman-Adv, *can* do seamless
>> handover of access point clients. But of course, there are lots of other
>> tradeoffs between layer-2 and layer-3 meshes.
> 
> I was thinking that if I am given an IP address unique for the mesh, and
> I could keep this whilst connected to any node, wouldn't it be possible
> to have packets originating from me be properly routed regardless of
> which node I am connected to?
> 
> I.e. if when connecting to a wifi node I generate my local IPv6 address
> incorporating my MAC address, which can be assumed to be unique. And
> when I am then walking to a second access point, losing connection with
> the first, I would resend whatever TCP packets not properly sent to the
> new AP, effectively re-establishing my connection on the new route.

The issue, as far as I understand it, is that your client won't
disconnect from the current AP and switch to another AP until the
connection gets so bad that it is forced to disconnect. The ideal
situation, on the other hand, would be that your client would switch to
another AP at the optimal moment, not when the connection to the first
AP gets so bad it drops. But unfortunately, that's how wifi works with
APs with different BSSIDs. It might be worth asking the Batman-Adv folks
how they do seamless handoff b/w APs.

> 
> Though I guess this is where I run into trouble in my train of thought,
> as it would perhaps mean that each and every route to each connected
> client would have to be propagated over the mesh network in order for
> the "return route" to my machine to be properly updated.
> 
> Sorry if I haven't fully understood low-level implementations of the OSI
> model, feel free to point me to documentation if it's too much. I'm
> mostly trying to walk myself through the complexities of mesh networking
> and asking for some help along the path ;)
> 
>> A while ago, I came upon a solution for doing seamless handover for
>> access point clients, developed by folks at the Distributed System and
>> Networks Lab @ Johns Hopkins. It's called Smesh: http://www.smesh.org/.
>> I would love to try to implement this in Commotion, but it would be a
>> long-term project that we currently don't have anyone working on.
> 
> No rush in my opinion at least. Seamless handoff is a luxury feature, so
> there's no rush in getting it working before anything else .)
> 
>> But keep in mind that we are currently ramping up development of
>> Commotion desktop/mobile mesh client software (linux, android, mac,
>> windows, iOS), that will let mobile phones and laptops and such become
>> full-fledged mesh nodes, which will allow for theoretically seamless
>> roaming, as they will use ad-hoc links instead of access point
> connections.
> 
> Absolutely, I'm following the development closely as we're trying out
> the firmware builds here in Umeå, Sweden. .)

Cool, keep us in the loop with your progress if you'd like, I'm
definitely interested in hearing about that. Good luck!

> 
> 
> 
> _______________________________________________
> Commotion-dev mailing list
> Commotion-dev at lists.chambana.net
> https://lists.chambana.net/mailman/listinfo/commotion-dev
> 


-- 
Dan Staples

Open Technology Institute
https://commotionwireless.net