[Commotion-dev] interesting tech offer from IRC

[Jeremy Lakeman]

Absolutely, I'd love to read it.

We've come to the same conclusions, that trying to assume that a
wireless network can be treated in the same way as a wired network is
a flawed model. It constrains the methods you can use to solve
problems. Also that using IP packets to sense network topology,
duplicates the work that is already going on at the wifi driver level.

I've been doing some almost pure R&D with a novel approach to link
state tracking and path selection. Though ATM I'm only trying to
determine the path back to one node, I eventually intend to try and
generalise some of the ideas.

But I've never simulated anything. I'm testing this in the real world,
with broken adhoc wifi implementations, primarily on android handsets.

On Fri, Dec 16, 2011 at 8:23 AM,  <advice at csmctmto.bizfu.net> wrote:
>> This is also something that The Serval Project is working on. Though
>> we're only in the early stages and have some other higher priorities
>> ATM.
>> We're planning to solve unique ids, spoofing and MITM by using
>> elliptic curve cryptography at the network layer. And we have plans to
>> load balance and store and forward some traffic types to improve
>> reliability.
>>
>> But we'd certainly welcome any other ideas you might wish to share. We
>> aren't naive enough to assume that there aren't going to be any better
>> approaches.
>
> My entire approach has been effectively at the network layer.  To really meet
> my original design criteria I had to design what amounted to an IP-equivalent
> protocol to manage transport, routing and link management.
>
> Is there anyone there who would want to get a five page braindump from me
> to try an initial implementation?  I can try to find my old notes and make
> them available, or describe the system anew.