Some technologies don't travel. Europe's superior cellphone network is one example and the spread of very cheap Wi-Fi in the US is another. In Europe, we pay more for Wi-Fi, but have better cell service. Authorities in the US have begun to use their cheap Wi-Fi to make good the problems with cell service, by putting together meshes that cover whole cities (see our group of case studies). .

The concept has yet to take off in Europe in a big way - perhaps mainly because it is not needed so much. Here mesh networks are beginning to be used to fill a different gap: poor provision of broadband. But as US mesh vendors arrive in Europe, it begins to look as if they are promoting yet a third Wi-Fi meshes application: as an option inside buildings.

Does mesh scale?
Wireless mesh is a simple concept: set up nodes across the area you want to network, give one a connection to a wired network and let the others use wireless links to connect up and cover the area.

But users are sceptical. If the mesh uses the same frequency for back-haul then the data the system can carry is severely limited. If all the bandwidth is being aggregated to one node, then the throughput of backhaul links close to that node will limit the total throughput at the edge of the mesh. If the same radio is used for backhaul as well as data access, then access to the network will get harder as each node spends more time forwarding traffic.

The debate about mesh scaling escalated (sorry) when Francis daCosta, founder and chief technical officer of mesh company MeshDynamics told a rather surprised wireless news site DailyWireless, that meshes don't scale. The rest of the wireless world leapt in and argued the other case, and Wi-Fi Network News took up the discussion.

The problem that emerged, after several contributions of varying complexity, is simple. There are currently so few mesh networks implemented that most of the ideas about their capacity and scalability are based on simulations, theories and a few networks built in lab environments.

The good news is that the backlash was way overstated. daCosta's initial contribution assumed that all nodes can see each other, and that the backhaul is on the same frequency range as the networking service. Both are obviously not necessarily true - most mesh networks now use multi-band nodes, with typically 802.11a used for backhaul for an 802.11b/g network.

Beyond that, the bandwidth that individual links will carry can be increased by using different media access control techniques (MACs). In the US, it is relatively easy to up the power of the links beyond the 500mW limit (Ofcom in the UK might be more restrictive).

Follow enough of the links and you get to academic papers that prove that, with a detailed enough theory, and a separate channel for backhaul, you can predict good throughput and good scaling with mesh networks. The next step is to transform that into reality.

So let's sell mesh
Plenty of alternative mesh architectures have been developed, and are in real-world deployments. The IEEE has a potential standard, < ahref="http://www.techworld.com/mobility/features/index.cfm?FeatureID=695">802.11s, which will add mesh as an option to Wi-Fi networks.

Among the vendors, Nortel is the biggest to back the technology, with vendors including MeshNetworks, BelAir Networks, Firetide, RoamAD, SkyPilot, Strix and Tropos. jockeying to get their technology installed.

Mesh for broadband, or mesh for buldings?
Given the right technology, some vendors - Tropos, for instance - reckon mesh can take on 3G. In the UK, LocustWorld is attempting to build grass roots support for its MeshBox systems, which can build broadband wireless networks.

Meanwhile, others are pushing mesh as a means to provide a network in hard-to-wire buildings. US-based Strix Systems, which has just opened for business in Europe, is best known for its presence in city-wide meshes such as Hermosa Beach, California, which claims to be the first "Wi-Fi city", However Strix also reckons it is a hot contender in the office.

Strix ran a contest earlier in 2004, to find and network "Ethernet unfriendly" buildings. When it is difficult to get cables around, or in areas like gardens, car-parks and large halls, the option to go wireless is a big plus with customers, it says.

The Strix systems are made from Lego-like modules that clip together. Each Strix stack includes a base power unit, with optional Ethernet links, an antenna on top, and up to three assorted radio units which can build a mesh. The units are designed to look good on a desk, but are plenum rated to go in ceilings, says marketing vice president, Doug Huemme.

Strix believes it has got the scaling problem sussed: "We have less than a five percent performance lose per hop," says Huemme. The bandwidth can be made even higher by plugging in more than one 802.11a unit in one stack, and using multiple channels for backhaul, he says, or by using channel bonding to get more data through one radio.

How will Europe go for it?
So far, city-wide Wi-Fi clouds are a mostly US phenomenon, where well-off city councils want to use connectivity for citizens to provide better service and boost business in their centres. In Europe, mesh has been more widely used by grassroots activists making up for the shortfall of broadband provision (see Locustworld).

The arrival of Strix looks like highlighting another Wi-Fi mesh opportunity: networking in old and awkward buildings. Europe has, shall we say, a greater wealth of such buildings than the US, so we expect to hear more of in-building mesh in future.