On Sat, Aug 04, 2012 at 10:59:09AM -0500, Jimmy Hess wrote: > On 8/4/12, Eugen Leitl <eu...@leitl.org> wrote: > > On Fri, Aug 03, 2012 at 08:31:06PM -0700, Owen DeLong wrote: > > onboard (as most smartphones and tablets do). > > 24 + 24 + 16 bits are just enough to represent > > a decent-resolution WGS84 position fix. Plus, > > GPS gives you a pretty accurate clock. > > Yes, very interesting. I wonder how do you achieve full scale > software testing for a mesh networking platform efficiently?
The Serval people do physical testing in the lab and the field. Of course, their scale is very small. You'd want to push most traffic through regular IPv6 Internet and only everything else through the mesh which doesn't have direct line of sight or fiber connectivity. This can be impemented as an abstract layer presenting a unified view, which uses VPN tunnels (each router node would not need to maintain many, even modest connectivites would result in very few hops) over IPv6 Internet just as Tor or I2P do it today. The penalty would be not that bad, given that you're not doing any deliberate traffic remixing/onion routing. You can prototype something like that quite easily based on CyanogenMod with IPv6, OpenVPN or tinc, gpsd, Serval, (maybe cjdns https://raw.github.com/cjdelisle/cjdns/master/rfcs/Whitepaper.md as well?) and some glue to tie it all together. > Do any of the virtual machine monitors Xen, KVM, etc > support an emulated 802.11n/other radio device that allows you to > configure "Emulated location and geography" for each virtual node, > to test various protocols and implementations across p2p wireless > meshes by simulating realistic connectivity performance? Very good point. I'm not aware of simulators which can do that on a very large scale (millions, billions to trillions of nodes).