Friday, May 27, 2005


Martin Geddes - Telepocalypse

At the risk of boring everyone to tears, some final thoughts on the emergency calling debate.
Firstly, who not just force telcos to unbundle 911 service? For those on DSL, they then pay $5 a month (or whatever) and the only number they can call is 911. The tricky bit is callback, but it doesn’t seem beyond all technological possibility to tackle this. The price might even be zero; you can call for emergency service from any cell phone, even if no longer connected to a plan.
I’m wondering if this routing/location thing is being made harder than really necessary. I mean, how difficult is it to have a system that goes to a human and says “National emergency service centre. Which town or city are you calling from? And which service do you require? Thank you, I’m now putting you through.” Yes, the caller might not know their exact location — but at least it approaches parity with traditional cellular 911 service (i.e. pre-E911 location enhancements). Also, self-reporting of location, or defaulting to a known subscriber or billing address is still better than nothing.
You could also have a directory of ISP routers and their locations — a less onerous task than listing sbscriber nodes. Geolocation services already have a lot of this sort of data. It’ll be incomplete, but a traceroute to the user will reveal at least some information as to which PSAP to route to. The VoIP/PSTN service provider could use this data in the absence of anything better. “We are now routing you to the Arlington, VA support center. If this is not the right location, press ‘1’ now for personal assistance.”
The physical routing of emergency calls isn’t as hard as is made out — but only as long as you put some of the burden on the connectivity provider, who knows where the end point is, and not the VoIP service provider, who doesn’t. For once, we can ignore the layered model and traditional good architecture practice and have as much layer bleed as we want. The whole source of the problem was the link layer, IP and session abstractions hiding essential physical information from the the application.
Why not just make every ISP route some fixed address like (think about it!) to a public service gateway? Or since DNS is virtually always bundled with internet access, just make a local DNS record that points to the right place (“ENUM ultralite”). Yes, by traditional standards it’s messy; but we don’t care, as long as it works well enough. There’s loads of ways of getting a connectivity provider to state somewhere the location information they know. (Some connectivity providers would have to confess to how poor a record they have of their plant and customer’s locations, mind you.)
Connectivity providers not using wireless links should, if they’ve done their homework, know the definitive address they laid the access to. Their data quality should significantly exceed that of a pure VoIP service provider, who has to accept the customer’s version.
The really big issue with nomadic IP emergency service is accountability. If voice calls become free, the service provider is cut out of the picture; at the most it’s like Skype where’s you’re an anonymous person with a self-assigned ID. You need a service provider mediating the system in order to have some comeback on abusers. It’s too hard to build a pure IP emergency calling system given the state of play with identity and reputation systems today. We don’t have any means of socially regulating a native IP emergency service offered via open Wi-Fi hotspots. You might use DNS or routing tricks to allow people to contact emergency service, but the system collapses from misuse.
Nonetheless, doesn’t anyone else find that an emergency service system that offers zero native interface to the Internet is, nonetheless, somewhat out of date?
UPDATE: See what I mean?


Post a Comment

<< Home