Last week I wrote a brief article noting how Korea is issuing phone numbers for VoIP applications, but encumbered with a number of restrictions on their use. David Isenberg has posted a comment that raises a billion dollar question: from a long-term view, is it a good thing that many VoIP applications are trying to re-use the numbering space of the PSTN? (I hope that’s a correct precis of your concern, David!)
This issue is so important I’ve decided to take a break from the endless task of unpacking our sea container and write a separate post about it.
So, is there a case to be made for re-using the PSTN address space for VoIP? I think the answer is a qualified “yes”. Nonetheless, a little voice inside me wants a cleaner break from the past.
Clearly VoIP technically doesn’t need PSTN phone numbers, as any user of Skype or Yahoo! chat will attest. But there’s a problem with trying to break free. VoIP doesn’t scale well socially and politically without a strong identity infrastructure.
The previous tying of the application identifier (phone numbers) to the connectivity provides a rough-and-ready means of providing accountability. For example, caller ID helps prevent and trace abuse. I wrote more about this in the summer. In brief, you can receive your teleconnectivity in four ways:
1. From a central authority to a fixed line into your own home.
2. From a central authority to a mobile device which is post-paid and for which you have passed a credit check.
3. From a central authority to a mobile device that is pre-paid, anonymous, but tied to one provisioned device.
4. From somewhere at the network edge (e.g. ad-hoc networks, meshes, open WiFi access points) that is anonymous and not tied to any device.
In the first three cases, there is what you might call “identity collateral” built into the supply of the connection. That means if you misbehave while using your connection, there’s the possibility of retribution by society. Case #1: Your end point has a unique network address. Someone can come to your house and arrest you; Case #2: You are provisioned to the network, and that’s tied to an account name. Someone knows who you are and can tap into your government IDs or bank account details and track you; Case #3, someone can de-provision your pre-paid device as punishment. Varying degrees of suffering, but a much greater degree that, say, having to set up yet another Hotmail account to run your scams and access MSN Messenger voice chat.
Think of it this way. If you dial 911, 999 or 112 and there isn’t an emergency, you should be punished. You’ve risked lives for no purpose. Any VoIP system that can’t hold you accountable is useless for this service and cannot be allowed to connect to emergency services. Likewise for a thousand other abusable services, from the local pizza delivery outfit to dating chat lines, in varying degrees.
Now, the following is so vital, I’m going to put it in large red bold. It’s the only large red bold ever posted on Telepocalypse, and I promise not to do it again unless something even larger, redder and bolder is being said. Which isn’t likely.
What’s the core purpose of a telco? A telco joins the physical world to the virtual world.
They do this in two ways. Firstly, they map physical points on the globe to network destination addresses (e.g. other people’s IP addresses or phone numbers). Secondly, and often forgotten, is that they give you a virtual address of your own and tie it to a physical place, person or device. [Sorry, I’ll go easy on the red and bold from now on.]
My hunch is that the future of end-user telecom is in the latter function, not the former. The retailing of bit passing is a loss-leader to other functions that map the physical to the virtual. Indeed, much of the cost of running a telco comes from the care and sales functions that are aligned with creating and maintaining identities. The cost of building networks is dropping fast, while the cost of managing the customer relationship is not.
Consider a typical large mobile carrier with 20m customers. Say the network costs about $15bn to build, and generates a similar amount of revenue a year. The network asset turns over once a year. The “cost per gross add”, i.e. the cost of customer acquisition, is about $300-400 per customer, making a total of $6-8bn. This customer data itself generates less than $100m of revenue from caller name exchange, syndication into directories, etc. So the data asset turns
over (order of magnitude) once a century. Hmmm. I know where I’d be putting my energy in growing my telco: expanding the number of transactions the data assets can take part in.
Incidentally, the arrival of m-commerce and proximity payments is a natural extension of this model. Make that #3 on the list — linking a device to a payment instrument. I believe telcos eventually will be very successful at this function.
Obviously, telcos are not the only ones in a position to provide digital identities backed by physical collateral. You can see employers, banks, goverments, etc. doing this too. But telcos are uniquely well positioned to do so for the base layer of general-purpose public identities. This is because the overwhelming majority of Internet connections are of types 1-3 above, and the telco automatically is custodian of your identity collateral. Furthermore, the phone number is already the de-facto standard identifier, even for many “data only” devices.
Without PSTN numbers, VoIP systems tend towards “calling cliques”: buddy lists of people whom you are happy to receive calls from. The social and regulatory infrastructure of PSTN numbers makes that scale to the level of the whole of society. David and Martin being happy to publish their Skype addresses isn’t the same as everyone being happy to do so.
I’ve an amusing T-shirt from a recent conference which adds “layer 8” and “layer 9” to the OSI stack, and labels them “financial” and “political”. We’re talking scalability at “layer 9” (political), not layers 1-7.
In a nutshell, the case for retaining PSTN numbering for VoIP rests on the ubiquitous existence of identity collateral in connectivity provision, and the pervasive use of PSTN numbering to represent that identity relationship. Whilst this is an accidental by-product of tightly coupled connectivity and service, and new identity system that fails to embody identity collateral is inferior and will not succeed in mass adoption.
The qualifiers? Firstly, the system of PSTN numbers, as we have both noted, may break down within a decade or less. Caller ID spoofing, number exhaust, geographical toll bypass, VoIP spam, regulatory by-pass to avoid associated universal service fees etc., the list goes on. Secondly, you can argue that the need for a single, monolithic real-time voice application called POTS has gone away, and there isn’t a need for a monolithic VoIP replacement with a single identity infrastructure. A balkanized voice world of varying degrees of central control and interoperability is the primodial soup of innovation we have to live with. Passing chaos is the price of change.
PSTN numbers also come with some unwelcome legacy. As my brother dryly put it today:
I see Skype switched your id across even though you not only moved house but also IP address provider.
So number portability is a crap imitation of what we need. Also the geo-centric nature of numbers and the consequent outrageous charges for crossing the boundaries is a barrier to keeping the system alive.
We could also see other systems emerging, where the focus on PSTN numbers is a distraction. For instance, let’s look at this entry in the RIPE database for IP addresses in Europe:
% This is the RIPE Whois secondary server.
% The objects are in RPSL format.
%
% Rights restricted by copyright.
% See http://www.ripe.net/db/copyright.html
inetnum: 82.70.155.144 - 82.70.155.151
netname: ZEN000021806
descr: Mr Martin Geddes
descr: ADSL
country: GB
[...and so on...]
Because I have a static IP block from my ISP, I’m in the database. But as a basis for an identity infrastructure it’s positively paleolithic compared to the PSTN’s numbering system. Too sparsely populated, too little asserted about the content, too fragmented governance, too opaque to the everyday user. But I’d love to be involved in fixing those problem. I just suspect it’s a generation’s work.
Another objection is that new forms of identity collateral may be required for new types of service. For instance, using your eBay login as the identifier and and your reputation score as collateral. Maybe by 2015 you’ll need to lodge some eGold in an escrow account to make a phone call and assuage fears you’re a VoIP spammer. (Only Warren Buffet will be able to affort to call Bill Gates!)
Make no mistake, identity issuance is fundamentally centralized. Why? Because we issue identities not only to distinguish between individuals, but also to regulate the behaviour of those individuals in a social and economic context. When you give details for a credit checked identity, lying can result in imprisonment. That coercive power is ultimately vested in governments, who in principle have a monopoly on the legitimate exercise of violence to enforce good behaviour. And unless you’re an anarchist, government authority is central in execution. (Please, don’t write in about the sovereignty of the masses. Please. Not here. Also, for those who believe corporations are usurping the power of government, I suggest you try being imprisoned, have the baliffs retrieve your goods or having your home suffer compulsory purchase before you go much further.)
So, finally back to the Koreans. The two steps forward are having tons of ultramegabroadband, and getting VoIP-PSTN numbering sorted. One step back for assuming the importance of PSTN numbers comes from mandatory technical rather than social characteristics.
Now, I wonder what’s in that mysterious unpacked box in the corner of the study…
UPDATE: Here’s another way of putting it, maybe a little more succinctly and coherently. Connectivity provision creates de-facto identities via assigned network addresses tied to account numbers. These are backed by identity collateral, by the nature of bridging the physical and virtual worlds. The application layer is wholly virtual. VoIP is just a software application. But to socially scale it also needs to access some form of identity infrastructure that ulimately links to physical entities. The most natural one to adopt is the one we already have in connectivity provision. And the telephone number is an acceptable identifier already in global use by normal people. The issuance, use and governance of IP addresses is even more screwed up. The question is how to extract the best of PSTN numbering and translate that into an IP world. If it proves impossible to retain the (identity collateral-backed) phone number baby while ditching the (subsidy-ridden anti-competitive telco) bathwater, then we’ll have to commit infanticide. But not yet.
UPDATE: Maybe this is the central undecided issue of VoIP: incremental vs. revolutionary change. I’ll be bearing this in mind at VON next week and will try to gauge the wind direction better.
Posted by Martin Geddes at 12:24 AMTrackBack URL for this entry:
http://www.telepocalypse.net/cgi-sys/cgiwrap/mgeddes/MT/mt-tb.cgi/310.
Listed below are links to weblogs that reference Number of the beast:
»
Numbering and Addressing in VoIP: Wag the Dog? from Aswath Weblog
Lately there have been many discussions and announcements on numbering and addressing in VoIP. The industry is fixated on using E.164 numbers as the addressing mechanism in VoIP even though it creates problems or recreates the PSTN business models; the...
[Read more]
Insightful post if perhaps a bit longwinded. ;-) (not that I don't like reading it)
Anyway, I don't see telephone #s going away. The fact of the matter is that billions of people know how to use them and that these numbers work in any language. You cannot say the same for textual addressing of things I think.
We shouldn't be trying to change the behavior of so many consumers, we should make the technology do more useful things for less money.
(The TN namespace will eventually be entirely in the IP world, with a minority of it still tied to the POTS network -- but that's just a technology issue and consumers could not care less.)
Posted by: at October 13, 2004 02:36 PM