It really isnt the case that some people will print anything Intel tells them. Im sure of it. It cant be true.
It is the case, however, that Intel did get a lot of free publicity for its “initiative” on wireless mesh networks last week.
If we are to believe Intel and its representative on the new IEEEE 802.11s committee, W. Steven Connor (he looks like a nice guy, so why not?), “At present, there are no standards for this.”
It sounds pretty watertight. A recent report from the Intel Developer Forum described mesh networks as “self-configuring systems” where each node relays messages to increase range and bandwidth. It quoted Conner, a wireless network architect at Intel and technical editor of the IEEEs 802.11s task group, authoritatively saying that there are no standards for this.
His colleague supported him. “Roxanne Gryder, marketing development manager in Intels Communications Technology group, said existing mesh-networking protocols are all proprietary,” reported Electronics Weekly.
“Intel wants a standard to help it address markets such as home and office networking, and for public safety applications,” the report said.
Well! Jolly good for Intel, and we can all rest easier in our beds.
But isnt there just a teeny problem with this?
Well, I hate to be rude when everybody else is cheering, but I know a mesh network. It is based on a widely accepted open standard called the AODV, or Ad hoc On-demand Distance Vector protocol.
Its based on open-source code written for Linux and using open-source drivers. It uses IEEE 802.11a/b/g standard hardware. Its not, in short, even slightly proprietary. Engineer Jon Anderson, who created this mesh, was asked, “Why wont Intel use your stuff, then?”
He replied: “Well, if my stuff is just an ADOV implementation, then they already have that as a standard. On the other hand, if they want to modify it for operational reasons, they need to do what Ive already done. That is, a massive amount of testing and research in the real world.”
It gets weird when you get into the details.
Intels proposed IEEE standard will mesh 20 to 25 nodes. They will be “largely static,” and one day, Intel says, the company may introduce the concept of “mesh portals,” devices that can link complete mesh networks to each other.
When? “We may have pre-11s standard versions by 2006,” Intel said.
By contrast, Andersons LocustWorld mesh is already two years old. It will link not just 20 but 50 or 100 mesh nodes, and they can be moving.
Each one can be moving, but usually just one or two are, as in a recent data-collection test that had one node in a microlite plane and several ground stations all linked to it simultaneously as it flew over. And it does clustering–which is everything the portal might one day be–already.
Its certainly true that there are dozens and dozens of charlatans out there selling mesh networking to venture capitalists. I have interviews with some of these guys; people who have yet to ship a single working network, who nonetheless are quoted as authorities on the subject of how much data you can ship across a mesh network.
Next page: Operating controls, and low-cost to boot!
: Operating Controls”> Meanwhile, with more than 500 working meshes around the globe, engineer Anderson has already has QoS (quality-of-service) controls set up on his low-cost standard PC hardware, so that every mesh access point you set up can be an Asterisk phone exchange, automatically linking to every other mesh network in the world over the Internet with free calls, but also enabling the creation of low-cost village sub-exchanges.
Intel faces severe challenges technically. According to www.incidents.org, the 11s standard “is a wonderful application of wireless technology, allowing organizations to cover large areas without significant investment in infrastructure. I predict security to be a problem here, since the design of a mesh network is the same as that of a man-in-the-middle attack.”
But this is not a prediction. Its history. Anderson worked out the threat three years ago, and solved it in 2002 with full cryptographic, certificated meshing–that is, all routing data and each IP address is signed.
He commented: “You cannot simulate this sort of research in the lab or on paper. No amount of theory will work. Ive been out on my bicycle for the past few years, actually installing and fixing mesh networks, and Im really not that impressed by people who say we need another IEEE standard, because we have one.”
Proprietary? No: You can download the mesh code and source from LocustWorld, install it on your own PC and make your own mesh access point.
Only small-scale? Hardly! Have a look at Andersons clustering PDF file (hell mail it to you if you ask!) and see how you can join boxes together and route over multiple media.
“You can make a three-radio meshbox by joining three ordinary meshboxes together,” Anderson said. “Any device can replace the radio, too. So, just plug in lasers, power-line adapters–even proprietary Intel meshes if you like.”
If you need more bandwidth, “just add another link–it aggregates the bandwidth. This technology gives us access to WiMax or any 802.11g device and so on, without needing to fiddle with drivers.”
Theres a simple, obvious explanation for this. Intel simply never heard about LocustWorld, right?
Wrong. As long ago as 2003, I wrote a widely read analysis of what Intel was doing with mesh networking. I wrote: “There is one mesh technology which has been creating [commercial] neighborhood networks for 18 months now; a new mesh is set up somewhere in the world every day using it, and it is fully Wi-Fi-compatible. Intel and Cisco have dismissed it as being nonstandard.”
No, that wasnt the first time theyd heard of it. In February 2003, I remarked: “It looks as if Intel has decided to ignore everything that has been done in linking wireless network nodes into a “mesh”–and has, nonetheless, impressed observers with its originality of research!”
Thats more than two years ago, for anybody in Santa Clara, Calif., who has lost their calendar.
Anderson himself has spoken with members of Connors IEEE group, who say he should be talking to them. He takes the (possibly uncharitable) view that if they want consulting services, they should pay him for them, rather than asking him to pay them a subscription and waste his time in Disneyland twice a year.
This would, obviously, be a good idea for Intel. If they are as far behind LocustWorld as Connors announcement reveals they are, they need that help.
LocustWorld has written to Connor himself–so have I, by the way, Steven, so do feel free to reply, any day before I collect my pension. And somehow, Connor still feels free to go on telling the world that the mesh he will come out with—perhaps in 2006, maybe in 2007, which links only 25 static nodes—will be an advance over the mesh that is globally available, installed, working and tested, and has been fully functional for well over two years.
I cant understand why Connor feels he can get away with this. Maybe, after all, it is because there are some people who will, honestly, print whatever Intel tells them?
Check out eWEEK.coms for the latest news, reviews and analysis on mobile and wireless computing.