[consume-routing] New mobile mesh approach

[Peter]

On 22nd April '02, Jon Anderson wrote:

> Erm.... That is *exactly* what this facilitates.
[...]
> To actually route traffic using my code would require an addition to
> the kernel to handle these routing tables. I'm currently working on
> that.

And here's the rub.
 
There's little joy in writing kernel code if it won't ever get in. The 
alternative, maintaining private patches for an open source kernel, is 
a non-trivial task IME. I suspect that 16 million routes will require 
surgery on most platforms. Full-feed BGP will vex most kernels and that 
is nothing in comparison to what you have in mind.

Your methods for route discovery/delivery are interesting. I'm not sure 
broadcast UDP is the best way to go. Without seeing the code or more 
details of the protocol & packet structure there's not much more I can 
add.

While I like some of your ideas, for me they are dead in the water 
until you declare your licence. As you are considering an open source 
approach I instinctively recommend the breadth of adoption a BSD'ish 
licence offers. If not, that's cool, but with the present situation and 
NDA requirement it's hardly relevant to consume and it's aims or this 
list.

-peter