[Olsr-users] OLSR in Android...a good idea? (and stuff)

[Mitar]

Hi!

On Fri, Sep 17, 2010 at 9:22 AM, Robert Keyes <(spam-protected)> wrote:
> One of the reasons I have stayed with IPv4 is I have an enormous amount of portable IPv4 space.

What do you mean by that?

> The tunnel can and should be encrypted.

No, it should not be. Trust me, we are already doing that and we are
now searching for a way to remove it. Because all this small devices
(phones, routers) are too slow for fast encrypted tunnels. Even server
gateways can become a bottleneck once you reach hundred megabit links
or more. So what now happened in our network is that CPU on routers
limits throughput over the network. Not WiFi, not gateway, not uplinks
(we have a lot of fiber, 20 Mbit/s symmetric to homes here in
Slovenia, not so cheap anymore, but still affordable) but CPU.

And you do not gain anything by that. We run open wireless networks,
anybody can sniff traffic. We use unsecure routing protocols, anybody
can reroute all traffic over him/her. So why would then we encrypt
traffic over fixed infrastructure?

> The other thing that can be done with these tunnels is to aggregate several upstreams.

We are also doing that. But still problem is that routing protocols do
not use it directly. It would be great if they would.

> with the tunneling I describe is to avoid censorship.

For that and overall security end-to-end encryption is the best (and
only way?). Or make a secure (trusted) HTTPS proxy in the network on
some server people can use. You do not need to encrypt tunnels.
Especially because keys will have to be public (you will probably use
open source and you will probably want people to setup their own
nodes, and even if you will have closed hardware key will be sooner or
later extracted).

> My current endpoint is a
> rented server in the US, which has some advantage in content laws, but
> disadvantages in others (US protects political speech, but extends
> 'intellectual property rights protection' much beyond other
> jurisdictions).

Yes. This is probably the future. To interconnect different gateways
in different jurisdictions and select the gateway in smart way.

But another approach is to use something like Tor and simply
semi-randomly pick one jurisdiction among all available. Maybe we do
not need to reinvent the wheel.

> One of the things I'd like to see, though this may be present already, is
> a truely 'virtual' wifi device. That is, connect to any number of devices
> as a client, AP, or adhoc device, with different MAC addresses as chosen,
> perhaps on different channels at the same time (802.11b channels. Since
> 802.11g already has a DSP capable of covering 3 802.11b channels, I figure
> it may be possible). This way, I could use muliple upstreams, aggregating
> them through my VPN Tunnel. It would also be really nice if all of these
> were automatically associated with, so as I am travelling, each open AP I
> come across is automatically used to set up a tunnel.

Good luck implementing all this. ;-)

Even just making ad-hoc operating properly with standardized protocol
in drivers seems a hard thing. Now to make it behave as you describe,
good luck. ;-)

> One of the other ideas I'd like to add is a virtual link of two mesh nodes
> through another node, particularly through an AP. Imagine that two mesh
> nodes can't see each other, and this causes mesh isolation as a result.
> But two nodes can see the same open AP...if they both associate with the
> AP, they can talk to each other and unite the meshes, without using any
> bit of the APs bandwidth.

And how they will know that they can see same open AP? ;-)

I think such cases are border-cases (and probably it is even illegal
to just connect to open node even if it is open; if this is by
consensus then it would be easier to convert that AP to real mesh
node). Much better is to concentrate on making existing mesh
technologies better. Nodes you can control, nodes you can configure.

> If you use any of my ideas, please attribute them to me. I may want to go
> after my master's degree and then perhaps doctorate and I'd love to write
> a thesis containing these ideas.

My personal opinion is that ideas are too valued in this world. Good,
effective and quality execution of them is what counts. Ideas are
cheap, everybody can have them, everybody can say they have them. What
is hard is to make them reality. And even more, it is hard to nearly
impossible to attribute everybody's ideas - which ideas have your
ideas been based on, maybe not directly but indirectly? This is how
new ideas evolve, on shoulders of previous ones. And this is what I
believe free culture is all about. Free sharing of ideas, techniques,
concepts... Yes, attribution is important, but not just for ideas, but
for good execution of those ideas, if you will make at least a working
proof of concept of any of those ideas... Because, to be frank, you
have not told really anything new. At least for me. New would be to
have this working in all might you described.

So please do not write thesis containing these ideas, but thesis
solving and implementing them. This would be useful.

(Maybe my opinion is biased by my (mis)fortune to be able (and unable
not to) almost all the time generate new ideas. What I learned is that
then it is important to pick the right one and get it through to the
end. This is what counts. Talk is cheap. And everybody can say that
they had the idea first. But maybe they just expressed it first.)

I am glad that you have ideas, but this is just first 5 %. Please take
it to the end. This is the only way how you will learn a lot and make
a contribution to the world.


Mitar