On Fri, 27 Feb 2004, Alex Bligh wrote:

Because you always want to get to an E911 service in the same AS number...

You do or you dont? I dont see why anycast addresses need or need not be restricted to same AS.

(seriously, read the sip & sipping w/gs)

Havnt got the time. :) Unless you have a URL or somesuch as to put me straight as to why anycast addresses would not be suitable. However, I'll wager that at some point a URI is translated to an IP address, be it by client or gateway. And at that point, the mapping can be to an anycast address.

Alex

regards, -- Paul Jakma paul@clubi.ie paul@jakma.org Key ID: 64A2FF6A warning: do not ever send email to spam@dishone.st Fortune: Got Mole problems? Call Avogadro at 6.02 x 10^23.

On Fri, 27 Feb 2004, Alex Bligh wrote:

Anycast topology tends to follow AS topology, as people prefer their own routes.

Indeed yes.

routes. If you take a rural situation where you have your nearest (geographically) E911 service on some long link into Sprint, and the customer on some long link into UUnet, it is most unlikely they will be close (network wise)

Indeed. But who said anything about having anycast addresses routed /that/ widely? The anycast address can simply be the well-known 'emergency service' address. It can then be up to each service provider to route calls from customers to the appropriate emergency service provider. Eg, in your example above, if we decide we do not want to put the onus on customers to configure their 911 service, then UUNet are in the best position, indeed noone else is in a better position, to know that $long-haul-customer is located in Ruralville, so UUNet's E911 'switch' would need to have a table of (customer or POP) -> (appropriate emergency service provider). UUNet can deploy as many or as few of these switches/gateways as they want. Ie, anycast == well-known-address, every VoIP phone could be pre-configured with it even. From users POV, it'd just work presuming the network provider does the needed work. Even for mobile users. And even if they dont, it would still allow for the provision of public emergency service switches/gateways as a fallback which have some chance of being on the same continent, occasionally even the same country, where a 'best guess' could be made (eg by looking to see whether the inetnum for the source has a country: attribute or by some other IP->locality mapping technology). This presumes SIP has the facilities to support the needed gatewaying/redirection.

Anycast is arguable good for finding the best *connected* (i.e. closest using a network metric) server,

Indeed, however you are presuming the servers are far and few between. It need not be.

need to. For E911, it does.

Yes of course. Difficult problem though. And the best place to decide is as close to the person making the call as possible.

Alex

regards, -- Paul Jakma paul@clubi.ie paul@jakma.org Key ID: 64A2FF6A warning: do not ever send email to spam@dishone.st Fortune: Use what talents you possess: the woods would be very silent if no birds sang there except those that sang best. -- Henry Van Dyke

Crist Clark wrote:

To steer a little ways back on topic, perhaps looking at the standards for how mobile phones deal with emergency services is better analogue for mobile IP phones than how POTS does things.

Install SRV records to the reverse zone to give you emergency, directory, etc. service numbers of the locality? Unless you are the rare beast with Mobile IP this would probably work alright in 99% of the cases. If you do NAT, you deserve your faith. Or ask your vendor to do an ALG for your outbound calls with telephone number remapping. Just don´t make me pay for it. Pete

Valdis.Kletnieks@vt.edu wrote:

20 years ago, 911 was able to say "unless you're the rare beast with a cell phone, basing it on the physical service address that the copper runs to would probably work alright in 99% of the cases".

Let's not make the same mistake again.

So all IP phones should be outside of buildings and equipped with GPS or Galileo receivers? Pete

Valdis.Kletnieks@vt.edu wrote:

On Fri, 27 Feb 2004 21:19:48 +0200, Petri Helenius said:

...

So all IP phones should be outside of buildings and equipped with GPS or Galileo receivers?

I can think of plenty of buildings where you'd want the GPS even inside if feasible. Think "any mall or office buil;ding over 250K square feet.".

Sure. I would also like to my car to be VTOL capable. I think the real question is that are the phone and network vendors willing to implement systems which give out your position only when you want or if the handset gives your position with every call and there is a filter in the network who gets to receive it. So far all positioning systems implement the worse option. But there will be people who willingly give out their position with inch precision if they are told that it will increase their security. Pete

At 02:49 PM 2/27/2004, Jeff Shultz wrote:

** Reply to message from Petri Helenius <pete@he.iki.fi> on Fri, 27 Feb 2004 21:19:48 +0200

...

Valdis.Kletnieks@vt.edu wrote:

...

20 years ago, 911 was able to say "unless you're the rare beast with a cell phone, basing it on the physical service address that the copper runs to would probably work alright in 99% of the cases".

Let's not make the same mistake again.

So all IP phones should be outside of buildings and equipped with GPS or Galileo receivers?

Pete

Does anyone actually offer a mobile IP phone service yet? Does anyone plan to?

With Vonage you have to tell them where you are located so they can set your 911 service up to the proper 911 center.

You can take your Vonage with you. Some people do this. It's a bad idea to dial 911 on a Vonage setup that you have on the road with you, but the question of mobility certainly can be studied using this service.

With cell phones it's based on the cell it comes into. If some sort of truly mobile IP based phone comes in, I'd guess that the provider is going to have to set it up to where the local router (or associated VOIP device) "listens" to the VOIP traffic for a 911 call, intercepts it and sends it to the local 911 center - my presumption is that they'll have to have a router of some sort in the local area to handle the mobile IP traffic. The GPS idea isn't a bad one either - since I think most new cell phones are coming out with this (it's been mandated, right?) it's a cheap addition and can be used by whatever the router redirects the call to for a better determination of the call center if the phone has the info.

The easier solution would probably be for the "mobile IP phone service" to set it up as a dynamic address thing, where the phone number is assigned to the MAC address and the system updates a central index of what IP address is currently serving what phone number. And by whatever "DHCP" server assigned the address, that would be used to determine the 911 center most appropriate.

Despite best efforts, there will be IP telephones in residences that are tied to corporate phone systems that are far away. If in the heat of the moment someone dials 911 on that phone instead of the "home" wireline phone, the 911 call could well be routed to another city, or even another continent. Remember that phones could be attached to VPNs going far away. The simple response to this situation is to say "tough luck. Use a wireline phone." Or at least "don't use the corporate phone that's on the VPN." Will these be acceptable to regulators? Will the courts accept the argument in a court case? This isn't going to be a simple issue to settle.

At 03:21 AM 2/28/2004, Adam Rothschild wrote:

On 2004-02-27-18:43:50, Daniel Senie <dts@senie.com> wrote: [...]

...

...

With Vonage you have to tell them where you are located so they can set your 911 service up to the proper 911 center.

You can take your Vonage with you. Some people do this. It's a bad idea to dial 911 on a Vonage setup that you have on the road with you, but the question of mobility certainly can be studied using this service. [...]

I'd take your argument a step further, and argue that it's a bad idea to blindly rely on any carrier, VoIP or otherwise, to provide 911 services, period. Test it periodically (and of course, make your intentions clear, and confirm that there is no real emergency to report), lest you'll find yourself in a position where you need to use it, but can't.

Testing 911 or equivalent on a periodic basis is actually a good way to get in trouble with local authorities. In at least some jurisdictions (the community I live in, for one), dialing 911 gets you a visit from the police no matter what. They are required to roll on all 911 calls.

Vonage's E911 implementation (which in actuality is outsourced to some provider whose name I forget offhand) is broken in too many ways to enumerate in this forum without totally losing it and committing multiple NANOG AUP infractions. Try it; chances are you'll be routed to your local coast guard platoon commander, or the highway patrol, or dead air, or just about anywhere other than your loca PSAP. Not to merely single out Vonage, I've had my fair share of encounters with other PRI- and VoIP-based CLEC's who are equally broken, or at best, will hand off your calls to the proper PSAP, but with blanks for all informational fields except your BTN.

Our 911 system is broken and vulnerable enough as-is. I question the business ethics of anyone who would want to add additional layers of complexity and inadequately tested technology (like... *cough* VoIP), risking life and limb in the process, and say it will be there for you when you need it. And, much as it pains me to say it, things aren't going to get better until the FCC comes down hard, and fines all the basement establishments with 2 BRI's and a mic^H^H^HAsterisk box who purport to offer [E]911 service and in fact... don't.

Finally, quick show of hands for you RBOC's and cable operators out there... how many of you can say with confidence you've got battery backup on all your DSL-enabled RT/DLC's? Cable headends? Active HFC distribution gear and amplifiers in the field? I rest my case.

It's just as bad on the wireline side. We've got a number of SLC-96 mux units on the POTS network here. The batteries in the units have gone bad, and not been replaced. When the power fails, the phones go dead. There's a dual-position knife switch under the pole-mounted SLC units here. The lower position is to connect to a portable generator which in theory the telco would position in the event of an extended power failure. However after an automobile accident that damaged one of these, the replacement cutoff switch no longer has the generator port. Coupled with the dead batteries, there's really nothing worthwhile in the way of power outage handling in the POTS system, any more than there is in other technologies. In recent power failures, the cable modems continued to work, but the Verizon phones did not. So, we had our Vonage line, but not the POTS lines. (We have an auto-start generator in addition to UPSs with good batteries, so we had power ourselves). Cell sites have backup power, but in many cases no generator. When the batteries go, the cell site drops out. The question in all cases is what is the level of service acceptable to regulators and emergency services coordinators? Clearly there are problems of both power and call routing which must be addressed. It's unlikely NANOG is the forum for specifying standards in this area. It is similarly unlikely the IETF is the appropriate body, though it may be a place to figure out how to meet the requirements specifications of some other body.

...

The simple response to this situation is to say "tough luck. Use a wireline phone."

Yes. Very much agreed.

You took that out of context. I was not making the argument that such a statement was sufficient. For those developing VOIP technologies, they are going to have to come up with a better answer. As you note, the 911 system has a lot of problems today. As I noted, the wireline network has the same power vulnerabilities as the VOIP and cable-based telephony services and wireless services.