I'm starting to wonder how this will work. If it takes 3 hops to get to the person I'm calling, doesn't that mean the two phones in between have to each carry 100% of the data of my viop call? What happens when my phone is carrying the data from 25 other calls? How do I check my email then?

Voice alone do not demand much bandwidth, but is very sensitive to latency. Checking email can be high bandwidth if some relative or colleague sends a binary file of some kind, but do not care much for latency. Still, i think the primary use is one of disaster communications. Like being able to call or message that one is safe or in need of help, and where one is located. Think of it like turning a phone into a digital CB radio, with the added ability to forward as needed.

Ability to send text messages in a large mesh-network at least would be useful and many would willingly join such a network because the data amounts relayed would be minimal. P2P GPG-encryption of the messages should be implemented also (compression comes along implicitly then). Multicast and broadcast text messages in some restricted way could be supported also.

What I was talking earlier, the mesh network like this should also be able to use GPRS/UMTS-data to relay then when next WLAN-hop is not found. Mesh-network participants could enable and disable the feature acting as a cellular-relay and maybe also select if they are willing to relay only text messages or voice also.

all those projects looks like utterly useless to me. Sorry about this, they are cool but there are some problems, like:
-in emergency, if communications go down is really likely for power to go down, too. If we have a 100 meters range in a place completely devoided of life, we have like 20 meters range in towns, that means we need everyone with power generators or their mobiles will run out of juice and you can say goodbye to the network.
-the wifi frequencies are not optimal for distance communications. What you would need to do for those in using something like Citizen's Band, that are free frequencies reserved for personal use. If we are able to get some decent bitrate on those frequencies, they will probably do much much better than common wifi. Ofc you need support in mobile's hardware.

Well I don't think utterly useless! Of course WiFi (ISM2400) was designed for short range communications. Yet in ideal circumstances it has been shown to work over one or two kilometers! If the power goes down the cell towers go down too (read... NO communications, including Ham Radio et al) but all the other electronic hardware like microwaves etc. are also gone (read...NO noise) and so likely some range improvement can be expected. Our batteries may last another 6 or 8 hours yet, time enough indeed to call emergency services, your broker, the wife or whomever providing their cellphones are capable. Therein lies the nub of the problem! Serval are developing a nifty solution that allows a phone running their software to act as an access-point (without the backhaul) and therefore be able to upload the Serval software to nearby phones. This combined with their MeshMS (read..SMS without the networks) could see the program propagated exponentially in short time. The system must work independently of the networks and it must run on our existing hardware...this is the challenge. If Serval save a single soul it will be utterly worthwhile!

Err, phone networks, and cell networks by extension, have been designed with a independent power source. Towers and backhaul have battery and/or generator backup to ensure operation during a power outage. And anything that can keep a wifi setup going can also keep a ham radio setup going.

any news on this project ? have you looked at android source ?

Regards