Slashdot Mirror
Working Toward Roaming For Wireless ISPs
hrhsoleil writes "In the category of: This seems like a no-brainer and why-didn't-someone-do-it-before,
according to SearchMobileComputing, the Internet Protocol Detail Record Organization (IPDR) is pushing a set of specifications that would allow users to roam among different providers' hot spots. IPDR is an industry group that addresses billing issues for wireless carriers -- they've got the Wi-Fi Alliance, Gric Communications, and the Canadian Wireless Telecommunications Association backing them up on this one so it might actually get off the ground. It's about time that wireless ISPs get their act together and make roaming possible. If I can go to almost any bank machine in the world and be able to use it without needing to sign up for a new account, why can't I do the same with hot spots?"
story here
I beleive the origional poster meant seamless intragration from one network to another, not just different access points on the same network. At this point TCP/IP doesn't do this very easily as the same IP number won't work on a different network.
Little Brother, watching the watchers
The lead into this article says the groups behind this standard are the Wi-Fi Alliance, Gric Communications, and the Canadian Wireless Telecommunications Association. The Wi-Fi Alliance has been unable to get traction under its branded Wi-Fi Zones program from venues that would rather just show the network they're part of; GRIC is the increasingly distant number 2 player in corporate aggregated resale (i.e., no hotspots, just reselling hotspots); and the Canadian group has very very few hotspots in Canada. The leading Canadian WISP, FatPort, isn't part of this proposal.
More likely, the GSM Association's roaming standards group that drafted a long document (referenced here in June 2003) on handling WISP roaming for hotspots (with members on the committee from some of the world's largest cell operators) will become the backend.
Or, iPass, GRIC's rival, which will gross about $200 million in 2003 after a very successful public offering this year, will make its clearinghouse standard, which requires standardized authentication, the de facto method of fee settlement and roaming across networks. iPass has 10,000 hotspots under contract now, including T-Mobile, Wayport, and other major networks worldwide.
Freelance tech journalist for the Economist, MIT Technology Review, Macworld, and others
You don't need a universally unique ID for the device.
What you need is a universally unique ID for the user. There's only one person with your email address. RADIUS realms uses the '@' to separate username from the realm. Since a realm is often the same as a domain anyway (although not always), this gives rise to an interesting idea.
Dialup ISPs have been doing limited roaming internally or among a limited number of ISPs partnered specifically for a larger roaming area for years. It's generally done with RADIUS using realms.
All a RADIUS server needs to do is to refer a request for a user in a realm it doesn't handle to the proper other RADIUS server, then forward back the response. Normally you must configure a RADIUS server with which other server is authoritative for which realm. There's no reason there couldn't be a TXT record in DNS that lists the authoritative RADIUS server for a realm that's the same as a domain name.
The other part is a bit more tricky -- the RADIUS server that is authoritative for the domain generally requires that the requesting device (an access server or another RADIUS server usually, but it could be a Linux box or whatever else that wants to speak RADIUS) be listed in advance, and that it shares a plaintext secret used for shared-key encryption.
Billing for usage-based access is often done straight from RADIUS login, logout, and traffic records anyway, so this part is easy.
What would need to be done is for public-key encryption to be used between devices (at least from oen RADIUS box to another or as an option -- it may be hard to get the firmware on certain access servers to do this) and for the authoritative RADIUS servers for one domain to be allowed to authenticate against another domain. With these fairly simple updates to the venerable protocol, it could allow universal roaming not just among dialups and among wireless ISPs, but even across those two types of entities. Then you still have the problem of getting deadbeat hotspot owners and ISP owners to pay for their roaming customers...
Note that cell companies don't all roam on everyone else's networks. There are a handful of networks, and there's coverage in most places by any particular carrier or at least one of their roaming partners. Some cell companies don't do roaming -- if you're off their network, you're out of luck.
So what's really needed is for ISPs and hotspot providers to sign mutual roaming contracts in the model adopted by the cell phone providers. Then, no changes to RADIUS would even be required.
It's often the fact that when you go to reinvent a wheel, it's simply because you didn't bother to see if that wheel existed already. This wheel's been in use a long time. Don't reinvent it.
Meanwhile, no one seems to be using the T-Mobile/Starbuck's WiFi service. I tried it. Beyond the expense, I was irritated that I had to log on and off using Internet Explorer, and there appeared to be no way to upgrade or downgrade your account without having to call the 800 number and trust your luck that the person answering the phone would actually do what you needed done. Without screwing it up.
I tried another for pay service in an independent cafe with pretty much the same results.
Face it, unless the government steps in and forces independent free wireless ISPs to charge a fee for service, I can't imagine for pay WISPs making enough money to survive, let alone roam.
At least in Seattle
"Under the spreading chestnut tree, I sold you and you sold me."
To solve all this stuff you need things to be addressed at multiple layers.
That's why the IEEE has started 802.21.
Evil people are out to get you.