Meshing aspects to consider:

1. Security. Your data is passing through everybody else's router, it is a hackers dream come true, so additional layers of security need to be added.

2. Proprietary IP.  There are many variants out there, some open source, but many come from companies that are building their IPO on that 'secret sauce' code. 

Many mesh solutions today are based loosely upon early DARPA approaches designed to solve speedy battlefield deployments. 

DARPA solutions focused on small data loads with a high degree of jamming resistance.  Consumers in the mesh arena are typically less mobile, are not being barrage jammed but do have substantially larger data amounts to transfer than is typical over a battle-net.

All these variations on a theme hold back adoption and of course thwart all attempts at interoperability. If your proprietary vendor fails, your entire network is then based upon a solution with no support, spares or future development.

And what happens when the IEEE 802.11s mesh standard becomes formalized in early 2007?

The Task Group, is quoted as stating that the group’s aim is to develop a standard wireless distribution system that operates between 802.11 access points and works with “stations that don’t know anything about mesh.”

Better choose a solution that is firmware upgradeable to support the standard.

3. Bad coding.  A number of the existing mesh algorithms have a fundamental flaw, they route using 'hop count' as the metric, and this usually means longer links, which require more EIRP (or a better quality receiver), and this leads to more co-interference, which reduces capacity.  The routing overhead will surely kill you as your network expands and hopefully so does your active user base.

4. Latency.  If you have a respectable amount of clients on your network, and they have four or more hops to get from client to Internet, the introduced latency and jitter created by CSMA contention windows will just kill any attempts at VoIP or live video transfer over about two hops, especially on single-radio solutions. And isn't VoIP the next big thing?

5. Single Radio Solutions. Some approaches incorporate a single radio unit that is tasked with both serving clients as well as meshing (pseudo backhaul) at the same time. There is a performance hit using this approach. 

Rajeev Chand, a senior equity analyst with Rutberg & Co., a "research-centric investment bank" in San Francisco was asked for his thoughts on mesh recently:

Q: Some people have downplayed mesh in the past. What's changing?

A: Some of us were wrong about this space. Two or three years ago we thought that mesh outdoor Wi-Fi would have technology performance issues that would prevent its applicability in key markets. What we have seen in the past two or three years is that, for the municipalities, cost rather than performance turns out to be the large value proposition. Even if there are four hops, they are just delighted with anything better than 10 kilobits per second, even if there is a little bit of latency. That in turn has created a very significant near-term market.

Open-Source Endeavors:

MIT Roofnet
Ad hoc On Demand Distance Vector (AODV) protocol
Mobile Mesh
Open Shortest Path First IGP
GNU Zebra
4G Mesh Cube
Champaign-Urbana Community Wireless Network

Commercial Mesh Vendors:

BelAir Networks 2.4 + 5 GHz operation
Strix Systems
Firetide 2.4 + 5 GHz versions
Locustworld
Tropos
Nortel Networks
Moto Mesh (Motorola) 802.11s draft compliant
Millennial Net
Telabria
Meshcom
Ultramesh
MeshDynamics
PacketHop
3e Technologies
SkyPilot
Telabria
RoamAD
DragonWave
Dust Networks
DeFacto Wireless AirMatrix
Sohoware AeroGuard MIMO 2.4 + 5 GHz operation
Hopling
Kiyon
Sensoria Corp
Cisco 1500 Mesh 2.4 + 5 GHz
Industrial Telemetry Inc.

If you want to experiment with mesh at home, then you could do worse than buy a few used Netgear WGT 634U's off eBay, load the impressively simple MIT RoofNet code and start running sustained throughput experiments etc. 

Cheers Nigel