Telepocalypse

Homo economicus

It's a common fallacy that all port blocking and IP blacklisting by ISPs is necessarily evil.

Two recent examples have shown up on Boing Boing. Now I love Boing Boing, and read it avidly. Sometimes, however, things stray a bit too far from being a Directory of Wonderful Things, and brilliant English literacy gets mistaken for economic literacy.

Both articles touched on ISP filtering. The first from Ecuador, translated into English here says:

"Satnet, one of the largest providers of Internet access in Ecuador ... has decided to censor net content, blocking access to any file-exchange services (P2P sites) ... In an act that goes against their own usage-policy, the provider has given different excuses for different users, spreading light on their poor customer service and useless policy regulations.

Satnet users, calling to ask for explanations, have heard different versions: [etc.]

This is painted as a censorship issue, but just as easily it could be seen as a bungled price discrimination effort. It's a simple consumer misrepresentation issue; you've been sold a service that's supposed to transport all packets to all ports, and got something less. They're trying to segment out heavy P2P users, who are getting more benefit and should be paying more.

On the other hand, consider this one complaining of a Canadian ISP banning home servers:

... It was reported in Vancouver that Canadian telecom giant Telus has outlawed home servers for its customers with residential highspeed service. Ports used by such ftp, telnet and IRC servers, among others, have been blocked. According to Telus, 'These security measures are designed to reduce illicit traffic.'

The excuse is clearly bogus, but you just have to recognise that this is again price discrimination.

Are these evil? That depends. If you've no choice of ISP and no choice of paying more to avoid restrictions, then your ability to partake in society is clearly dented if your information pipe is leaking bits into the soil. That's a public policy problem. If you can pay more, but have no choice of supplier, then there's probably a competition problem. But expecting suppliers not to price discriminate is a bit like hoping water will start flowing uphill. Finally, if there's plenty of supplier competition, and still price discrimination, then hurrah! The market is being extended downwards to people who otherwise could not afford to participate.

So without context, these articles that present price discriminating activity as a priori a bad thing are just meaningless.

Unless, of course, you believe broadband is a posiive human right and everyone should have unfettered access for free paid for by the goblins and fairies.

Not so super nodes

David Beckemeyer has some thoughts on Skype's terms on service and that most users are unaware that their PCs might become supernodes and relay other people's traffic.

There's a potential PR disaster for Skype looming here. A lot of people are on metered broadband -- mostly low-end products that extend the market for DSL to people not willing to pay $45/month for unmetered access. If your PC becomes a supernode, you might be paying hefty overage charges, or get capped. But if Skype lets people opt out of being a supernode, everyone will do it and the system collapses.

Funnily enough, much of the answer comes from David's next entry: stop needing supernodes by not hiding behind a NAT box.

You can see the telco marketing pitch now -- don't use Skype, use our "managed" VoIP and get a predictable connectivity bill each month.

Pimp this

February 28th -- the last day for cheapskates to buy their way into Freedom to Connect.

Reasons to go:
1. I'll be there.

2. It's the most important thing, ever.

3. Err...

4. That's it.

Beach bum

The family is in bed, and I should be too. It's hot here in this vacation apartment. "Air conditioning in every room" turned out to mean "noisy, unreliable aircon unit in bedroom only." I've ripped the cord out of the bottom of the phone in the bedroom, used a socket doubler in my laptop bag to improvise an extension splicer, attaching my travel phone cable, which just sneaks out from under the bedroom door into the kitchen with the added reach. On dial-up -- oh, it's just like the good 'ol times (not). But I'm willing to endure this hardship for you, dear reader. (You're still reading this, aren't you Mum?)

Anyhow, here in Barbados there are four licensed mobile operators, three with operating networks. Some even have groovy names, like bMobile. Competition is real, and there are plenty of deals being plied on the car radio. (But then again, one station keeps emphasising that stereo sound is what makes them so great, so some things are just a bit behind the times here, too.)

Landline calls within the island are free, with Cable & Wireless having a traditional stranglehold over long-distance telephony at extortionate rates. C&W still have a lot of influence, but things are being opened up slowly. Most locals make few international calls, and it's served as a very efficient way of segmenting the market and doing price discrimination.

The mobile operators have replicated the same tariff structure, with high off-island rates.

Now broadband is slowly seeping out, although prices are high. Want 1mbit DSL? Yours for USD166 a month. But the cheaper low-speed packages would do Skype just fine, and could be easily justified as a toll-bypass for C&W's usurious PSTN rates. Broadband rates appear to be based on unmetered packages with no differential pricing for on-island and off-island data.

Where things get interesting is what happens when toll bypass using VoIP becomes common, and how price discrimination for IP networks might evolve. In a competetive market like mobile telephony, price discrimination serves to extend markets downwards into otherwise unservable market segments. (Because if everyone could pay that low price, the system couldn't pay for itself. In uncompetetive markets, price discrimination merely serves to transfer wealth from consumers to producers.) Some people in Barabdos -- many of whom are monetarily poor -- only have phone service because of price discrimination and political positioning by C&W. C&W pretends to benevolently provide more universal affordable service in return for a stranglehold on international traffic.

As VoIP encroaches, the pact may fall apart. (C&W are being fully exposed to competition over the next year or so, anyway.) This will cause political upset. If someone isn't being over-charged any more, there's no money in the pot to curry poliical favours by offering cheap basic telephony. Or is there?

You could still price discriminate IP traffic into on-island and off-island. Give basic access to low-end users at cheap rates, but cap or surcharge off-island IP traffic. The only question in my mind is "can you make it stick?" -- is it too easy for some users to relay traffic on behalf of others in ways that are effectively untraceable. After all, you could set up your router to only see on-island IP addresses, some of which will be Skype nodes of people with cheap international access, and you can work out the rest.

Another wildcard is that it isn't hard to cover the more populated bits of Barbados with wireless broadband. Can you make your pricing regime stick if someone else can bypass the copper? I've seen quite a few C&W trucks out mending copper lines slung through the sugar cane fields. You only need one direct hit by a hurricane and suddenly wireless looks like a very attractive proposition.

These Caribbean islands have had a very stable telecom monopoly position, but are rapidly becoming laboratories in which the future of telecom gets played out at high speed. The sheer number of islands and different governments ensures diversity of experiments. We're likely to see both pricing and technology innovation here. It's one thing to propose covering the USA with wall-to-wall Flarion, quite another on 166 sq miles of coral outcrop.

Maybe I should take a break from the beach and hunt down some local telecom execs to interview? Just don't tell my wife. I might have my laptop privileges rescinded.

Skype segments

I just wonder if Skype could actually be a mobile carrier's best friend.

The problem they have today is that bucket sizes are increasing, and data differentiation isn't really working very well. Somehow they want to value price that network traffic, but how?

Well, take Stuart Henshall's idea of offering Skype interconnect from traditional cell phones by using dial prefixes. Since Skype users are mostly consumers and small businesses, you've got a ready-made way of segmenting out the consumer-to-consumer calls from the B2B and C2B ones. Offer unlimited Skype calls, a small expensive bucket for business calls, and voila! Profit. You might even find ways to tweak your network to ensure the Skype calls get 2nd-class status, and "business class" calls don't suffer blocks and drops.

BT Miscommunicator

Mea culpa, mea culpa. I take it all back. My review of the BT Communicator product missed one thing.

It sucks like the Old Sow of Quoddy.

Why? I tried calling Barbados today. (I'm opening a temporary Telepocalypse Ltd. office on the beach for a fortnight.) Call barred. Click help to find out more. You need Internet Call Waiting to take control over your calls. Available free to landline subscribers to our Call Barring Option (I don't have it), yours for only GBP1.75 a month (~USD3.30).

What!?!?! You want me to pay for you to unfuckup your system!

Let's just say SkypeOut did a perfectly adequeate job, and BT won't be getting such nice reviews in future.

Protocol mismatch

I'm at the home of my parents-in-law in Vilnius, doing some business expenses. The rest of the family are round at my brother-in-law's house. To avoid spreading my sneezes and diseases, I'm staying home alone.

The phone rings. Uh-oh. No caller ID display on this phone. Is it my wife? Some random caller whose speech will be encrypted in Lithuanian?

Unfortunately the PSTN will never be able to reconcile my Accept-Language: en with the caller's Content-Language: lt.

That said, Skype only makes a passing note of your language preference in your profile, only accepts one language option, and doesn't do anything with the data like warn you the person you're calling isn't likely to understand a word you say.

Interview: Dmitry Goroshevsky, CEO, Popular Telephony

This is the first of what I hope to be a series of occasional interviews with the “people who matter”: those that are disrupting, being disrupted, or might control the speed of that disruption. Dmitry was kind enough to give me an hour one Friday evening for me to try to probe a bit more deeply into Popular Telephony and the – legendary or mythical? – Peerio peer-to-peer telephony product.

Some basics to put the conversation in context. Popular Telephony is the company. Peerio is both the product and a technology. The technology manages the storage and interconnection of peer-to-peer nodes on a network. Think of how Java Beans provided a standard set of APIs and execution environment to single-node computing, and generalise to P2P. In particular, Peerio’s design assumes no central servers. Storage is distributed. All management and trust functions assume distributed nodes and data. The technology is being embedded in phones, and VoIP is the first commercial application being deployed on top of the basic Peerio technology.

Popular Telephony is also promoting GNUP. This is client software that runs on Peerio nodes and performs two functions. Firstly, it includes various media gateways that run on the client to bridge different communications systems (e.g. PSTN gateway, Peerio, SIP, Skype). Secondly, it’s a distributed directory service that can also be used as the “holy grail” single point of contact by extending the PSTN numbering space; someone calls your GNUP number, and it gets patched through to your Skype service.

If it works, and if the customers buy it, then the scope for disruption via a fully distributed telephony system can hardly be overstated. If Skype was subversive, this is positively treacherous as far as legacy telcos are concerned.

Many of you will want to know: is it real, and can they deliver? Well, this isn’t a product review. You’ll have to pay someone to do due diligence! My goal is to explore the potential significance of the technology, and test its boundaries and plausibility. I'm assuming the technology works as advertised.

The format I’m using for these interviews is to reproduce a lightly edited, but largely verbatim, copy of the transcript. I’ll post up some commentary and analysis at a later date. Podcasts? I don’t like the sound of my own voice, and although I recorded this one (just for note taking), the technology sucked. Maybe one day.

Read, share and enjoy!

Telepocalypse: Willie Sutton was a famous bank robber who was once asked “Why do you rob banks?”, to which he said: “Well, that’s where the money is.” So a softball question for you first is which markets are you going after and why?

Dmitry Goroshevsky: Enterprise, and the answer is exactly what Willie Sutton said.

T: Enterprise is a very big world; different geographies, different parts of the enterprise. How have you honed your focus?


DG: Is Microsoft Office going after any particular segment of the market?

T: So you’re saying everyone needs to talk just like everyone needs to write, yes?

DG: Yes, if you take into account Peerio is about more than just communications.

T: You’re pretty busy at the moment in Asia [forming alliances]. What drove you towards Asia?

DG: A significant market. Strategically positioning ourselves equally in three markets: Europe, USA and Asia. Asia is the second market we went to. Asia is a known early adopter in the cutting-edge technology market, particularly Japan. Somebody told me Japan accounts for 70% of the new technology markets in the world. We have a cutting-edge technology breakthrough, and we see the Japanese are ready to adopt these kinds of revolutionary concepts.

T: The users don’t care if it is P2P or whatever; they just want to communicate. They don’t care what the network architecture is...

DG: ...Not necessarily true. There is an awareness what is behind, and particularly in the enterprise market where they want to know what’s going on in their network. And when we’re talking about the Japanese, they want to know how it works.

T: What unfilled customer need would you say the Peerio technology satisfies?

DG: It’s like asking what need does the first commercial airplane fulfil. You could do it the old way, in cars, trains and ships. It’s just faster, better and cheaper. It’s the way we will communicate in the future. In the next 30 years we will have computers that, according to Hans Moravec, will be as powerful as a human brain. These computers will cost less than $1000. So, imagine: we will have a Fibre To The Home network with these powerful computers. How can this be served by a central device? If you apply common logic it’s just impossible, because the server side will need to mirror the storage and processing capacity of the edge.

T: So you’re anticipating the arrival of diseconomies of scale of centralised computing?

DG: It’s impossible, not difficult. The network will crash!

T: So the purpose of Peerio is not just to do something that’s 99% cheaper, but to do something a centralised architecture will never achieve?

DG: Exactly. Cheapness just happens to be one characteristic of it, but it isn’t the whole vision behind it. What we’re saying is that “serverless” is the only approach we will ultimately have.

T: So isn’t Popular Telephony really a distributed database or distributed computing company masquerading as a VoIP play?

DG: [laughs] We’re not masqueraders – we’re telling everyone who asks the right questions. We’re doing VoIP because it can drive revenue to us, and it’s a fast-growing market. But are you going to only do VoIP – NO! It’s about communications. It’s a network.

T: So what do you see as the expansion of the communications space that you’ll fill? What’s your roadmap?

DG: We have a roadmap, but I can’t talk about that or the specifics. We’re trying to shorten the time between the announcement and the availability of the product. I can tell you the internal goal is to release a new application every month. No promises!

T: You’re aiming at an enterprise telco manager audience. But doesn’t your product put them out of a job? They no longer have an empire of people managing complex PBXs and session border controllers any more.

DG: [rhetorical voice] Did it ever happen before? Have we ever had new technology come and eliminate thousands of jobs? Not really, it always creates more than it eliminates.

T: But they’re not the same people. The PC revolution kicked off with people in the accounts department wanting a spreadsheet and buying a PC on discretionary spending allowances of senior managers.

T: What’s the purchasing model for Peerio? Will people be able to pop down to Costco, buy a 10-pack of Peerio phones and plug them in, and not have to worry about the telco manager?

DG: Absolutely! But we believe we will open a whole new market for serverless applications on top of serverless networks. That market can potentially be a multibillion dollar market.

T: Again, what you’re really deploying is a grid fabric, and you happen to be trojaning it inside a phone.

DG: Grid is not a good word because it’s a different architecture, but yes. But we are deploying a serverless network on which we believe you can deploy any type of networking application. Some applications we won’t even have thought of right now.

T: What barriers to adoption are there? Is someone buying the first Peerio-powered device a bit like the first person who bought a fax machine? There’s this beautiful future we can paint where everyone has these devices, but how do we get from here to there?

DG: Because it’s cheaper and better. It has features and functionality you cannot buy for this money, or sometimes for any money, on the traditional systems. Siemens have talked for years about new applications on top of their PBX systems. They claim “you’re paying even more for my PBX, but it gives you the productivity”. I can tell you that on top of Peerio this “productivity” is just a joke. And it costs 95% less. For me, it’s a very easy answer. There may be some psychological barriers, but not really technological.

T: Let’s talk about those psychological barriers for a moment. If there weren’t any barriers then a year from now you’ll be a multibillionaire and every enterprise in the world will have thrown out all its phones and installed Peerio and fired every telco manager. There clearly are barriers and friction: where does it come from?

DG: We have psychological barriers for anything unknown; that’s our nature. But we also have a drive to adopt new things called “progress”, and that’s also in our nature. There will be people stuck in the previous universe; there always will be, and always are. There are also people who can clearly see the advantages, including the cost and functional advantages.

T: If your product is two orders of magnitude cheaper than what’s out there today, will you be looking to distribute it via standard retail channels, or will you need an expensive sales force to you out and persuade people to buy 10,000 of these phones for their enterprise?

DG: No, we want to do what Microsoft did with different kinds of channels and distribution mechanisms. They’ll be cheap because they don’t need to be expensive. We will have a Dell type of marketplace on our website. You there, pick you phones, select a gateway (or if you don’t need a gateway, GNUP), it’s shipped to you overnight via FedEx, and there you go – a telephony system.

T: Is the gateway your secret server?

DG: No, we can do it with or without a gateway. With Peerio you don’t need an expensive gateway, because Peerio provides the redundancy. You don’t need any special form factors or fancy boxes. Peerio provides the five-9s reliability, even if you just store it in your shoe box.

T: Should users be worried about the Peerio tax,? This is like the Microsoft Windows tax, which is because of the very strong network effect of Windows. You get very high margins compared to the cost. Once everyone has a Peerio phone, is there a danger of being dinged down the road by the Peerio tax?

DG: I believe in capitalism as a system, and market forces. I don’t like Windows and don’t use it myself, but that’s not because it costs $100. It’s because it’s worse than MacOS X that I’m using. Windows was adopted by all these users not because somebody imposed it on them. It actually was the best system around. It was the cheapest system around. As long as it was right, it happened. That’s what created Microsoft, and there’s nothing wrong with that. But if Microsoft try to impose anything, user have Linux and Apple. The market will determine our revenue. If we’re good enough and ahead of everybody else, why should people be paying us. It’s only a fraction of the fee they’re paying to the Siemens, Avayas, Nortels and Ciscos of this world. Why worry about our domination if the market is already dominated by five players, who are making enormous amounts of money today?

T: Enterprises are like islands, looking to be joined up. How will people use Peerio between enterprises? What’s required?

DG: Either as a configuration issue, or on top of a bigger Peerio network, for example GNUP. Enterprises create their own Peerio networks. You can create as many networks as you like.

T: How do you peer Peerio networks?

DG: You need a device joining the two networks which is identified by a special identifier. So you just need the rights to access the network via a special identifier.

T: These special devices, are they like supernodes [a Skype construct] inside the enterprise?

DG: It’s just any device that runs the Peerio software. Telephone set, Wi-Fi phone, PDA or computer. It needs to be connected to IP and the Peerio service needs to exist on the IP network. Thousands of Peerio networks can run on the same IP network concurrently, and you will join the one you have an identifier for.

T: So what’s the user experience? I’m at XYZ Corp., and I want to speak to my supplier at ABC Corp.; how does it work?

DG: Firstly, Peerio has replaced your enterprise PBX. You have your own Peerio network. Your enterprise might have hundreds of sites, but they are configured to act as one Peerio network. You will have a numbering plan proprietary to your enterprise across all locations. To call outside there are three different ways. The example you picked it the most esoteric. Firstly you can use any PSTN gateway you own. Or you can use GNUP to ask other service providers that interconnect networks to get you out. [Popular Telephony have an alliance with Stealth – Ed] This delivers the Peerio identifier, which isn’t the same as the number on the local network. If it’s the same – which is what Skype is trying to do for the enterprise, and I don’t understand – it’s one network everyone needs to be connected to. That gives you problems with security and reliability.

T: So Skype’s problem is a lack of private namespaces?

DG: Exactly. In our system you run your own Peerio network, and join other Peerio networks as a special administration task. You authorise a handset to join another network, from which it receives an identifier, and it can receive calls from the other network. It’s modular. So just as an example, if you have another company that you work with a lot, you can create another set of numbers you use just to communicate between your two companies.

T: We’ve talked about calling within an enterprise, and federating enterprises. The PSTN lets anyone talk to anyone else.

DG: That’s why we created GNUP, to allow you to interconnect.

T: What’s stopping someone using a GNUP number to send out telephone spam and not be traceable in any way?

DG: What’s stopping them today? The Peerio network is completely manageable and traceable. We can provide a special authorisation number for the authorities if they want to trace a number. The reason for that isn’t spam; it’s the terrorist threat. From my perspective that’s way more important than spam. Peerio is completely manageable, unlike some others. Although completely distributed, with no servers, with the right administrator authority you can have CDRs and malicious call trace and all the PBX features. If a government agency like the FBI formally asks us for an authorisation code to trace a malicious call, spam or terrorist threat, we will provide them with the special administration number to trace calls. The network is completely aware of the needs of the government and law enforcement. I don’t have a problem with people knowing this; the Peerio system isn’t there for people to be untraceable and do whatever they want [anonymously] in terms of law enforcement. Inside the enterprise, nobody can trace it except the system administrator. We are providing full ACD [Automated Call Distribution] capability to monitor and supervise calls.

T: GNUP provides a distributed directory capability. Is this a viable stand-alone business, or are you only in the directory business to enable the basic Peerio technology service? Is there a business model for directories?

DG: We sell Peerio. That’s our business. It’s an enabling technology. It includes a database in which you can store anything you want, including directory services. So an independent developer can come to us, buy the Peerio SDK for $1500 (or whatever) and do whatever he wants to, including a directory service. But there’s no business model for [directories] and in particular we aren’t charging any money for residential offerings like GNUP.

T: Do you think there is anything from the PSTN model that will still require charges [in VoIP]?

DG: No.

T: Should a system that is resistant to spam and abuse be charged for?

DG: The application developer can do whatever he likes. If he wants to charge for incoming calls, he will do it. The difference is we don’t take anything from the end user. We don’t have any of our business model based on the price someone is paying over the GNUP network to somebody else.

T: Do you think there will be one VoIP architecture that will bind all enterprises and consumers, or do you think we’ll see a fragmented world where for B2B people use Peerio, B2C is PSTN and C2C they use Skype?

DG: Speculation; Commander Data would say “Not sufficient information”.

T: Too early to tell, or question not meaningful?

DG: Question is pretty meaningful, and it’s too early to tell. The answer you’ll get from people to that question will more depend on what they like to hear and do. My futuristic view? There will be no PSTN, and also no other centralised-based services. Basically, nobody except Peerio has a valid model – today. But that of course doesn’t mean that there will be only Peerio. But it will be serverless only, otherwise it just won’t work.

T: So would you see ENUM, as a centralised phenomenon, as a dead-end?

DG: Of course! Same logic; devices at the edge are becoming more powerful and intelligent. The numbering plan is just a small piece of what they will demand. Why would you create such an infrastructure of huge machines to process this data on behalf of extremely powerful machines at the edge? You just cannot do that. Or at least you can, but it’s absurd. If you can do it without [centralised servers] why would you do it with?

T: But isn’t some of the data, like who does this IP address really belong to, centrally owned? If there’s value in that data, won’t they seek a central point of control?

DG: They are nothing to do with each other. Peerio provides you with full control without the centralised architecture.

T: But if the data is centrally owned, and a dip is not too expensive, and the connectivity is really powerful, who cares whether the other node is “centralised” or just a remote peer?

DG: The architecture won’t hold.

T: You’re betting that the volume of transactions will exceed what any centralised architecture can support.

DG: Of course.

T: We see the Oracle’s of the world out building their massive grid servers, with each server having tens of processors. Is this the diplodocus of the dinosaur era: really big, really hungry, and about to be really dead?

DG: Right. The scalability of today’s databases and transaction servers is an issue. Why won’t we solve this with a serverless approach? You just install Peerio on servers running transactions.

T: If your distributed storage technology is so good, why are you focused on telephony? Why not sell the technology into some other distribution channel? Why not sell it to Sun and get them to re-invent their business. [Yes, I get the irony - Ed]

DG: Just a question of time, effort, money and focus. Yes, we’re starting with telephony. Can’t do everything at once!

T: Thank you for your time.

Costs and benefits

Former AT&T exec Tom Evslin lucidly notes many of potential features that VoIP could introduce that would never appear on the PSTN. The success of a stupid network solution is about feature innovation, not low cost.

I think he's got one third of the picture right, but it needs the other two thirds to put it in context.

I wonder if we'll see three waves to the adoption of VoIP. The first is predicated on price and arbitrage. Vonage falls squarely into this category. The second wave is on features that really can't be done on the PSTN. Skype has a foot in both these camps. The third is social; people buy the device or service not because of intrinsic value of the features, but because of what it says about you. It's like not having e-mail. In 1996 you could reasonably have argued whether you needed it. In 2005 it means you're hopelessly out of touch, desperately poor, or purposefully eccentric.

The difference between the feature adopters and the social adopters maps roughly onto Geoffrey Moore's early majority and late majority. The late majority buy in not because they see the benefit, but the costs of not joining start to increase.

An example of social adoption might be radio call-ins; not having a wideband audio codec marks you out as a connectivity second-class citizen. People who sound good will, rightly or wrongly, be assumed to have more valuable opinions. It's the radio equivalent of body language.

So that's the fuller benefit picture -- it's more than just features. But we haven't really looked at the cost side. This is often neglected. People will adopt VoIP in large numbers when the benefits exceed the costs. So what are the costs? Nobody wants to talk about this because the dirty little secret of VoIP is that it can be expensive.

The "free postage and packing" for your voice packets that the Internet provides tends to make people think VoIP is free. That's only true if you narrow your view to the marginal money cost of the conversation. But most of the cost isn't associated with the call per se. First you buy smart hardware, and have to learn the new system. As long as the market is in rapid flux there's significant risk you're learning a dud technology. Then when it comes to the call itself there are mental costs; does this recipient have VoIP? Which client? Which version -- oh, and does the Mac version they have support this voicemail feature? Will your Siemens IP phone interoperate with my Panasonic one? Will my privacy be protected? Will I have to tell everyone my new contact address?

So the whole transaction cost of a VoIP call is way more than just the bit haulage. Admittedly the new features from VoIP are dropping the transaction costs of some calls. I find it much easier to co-ordinate establishing a conversation using voice messaging, IM and presence built into a tool like Skype. The "cost" of a successful interaction on Skype is lower than playing PSTN telephone tag.

Skype's integrated proprietary architecture, and focus on usability, can be thought of as a way of minimising transaction costs. That "no spyware" tag on their front page is important in lowering the investigation costs of the prospective user. You also don't have to worry about interoperability of different flavours of Skype. (Their rejection of an open SIP architecture was a rational one.) The "total cost of ownership" of Skype is low. Your mum can install it without you being there to help. You don't even need to be aware of the NAT are firewall issues, let alone solve them yourself.

The familiarity of the PSTN gives it very low mental transaction costs. The user hardware is also pretty damn cheap. Where competition and (de)regulation work, prices are low. Its extraordinary tenacity in the face of the IP onslaught is testament to its success in lowering total cost, despite high headline call charges.

The achilles heel might be voice spam, which increases the cost of ownership of a PSTN phone. Smart, automated and malicious edge nodes weren't anticipated in the PSTN design. We're all trustworthy telcos, and the users can't do much harm to each other with the dumb devices. Bridging the Internet to the PSTN ruptures its trust model. It's a sudden environmental change of the type maladaptive disosaurs dread.

Should this get to be a big problem, IP telephony will be adopted more rapidly by the masses; but not because of new IP features but because of a changing cost-benefit equation compared to the PSTN substitute. Diverse and distributed IP telephony will offer more means of resisting spam than the PSTN monoculure. It's the parasites, predators and diseases that drive rapid evolution. We didn't sprout ears and eyes because it was cool, but because it helped our small furry ancestors avoid getting eaten.

The summary is that new forms of IP telephony will be adopted for richer reasons than just features, and the adoption is driven by a more complex cost-benefit landscape, which includes the changing true costs of participating in the PSTN (previously down, tomorrow up).

CANdo notes

Earlier I promised to post up my notes from the Community Networking get-together in the UK on 19 January. I've cleaned them up, so here they are. I'll post some commentary another day.

Malcolm Corbett

Gave a summary of findings of the survey of community-based networks in the UK. [Details will be publicly released in March.]

Found projects active in over 500 locations, mainly wireless, mainly new. This is more than expected and CBN does not believe that all projects & activities have been located. Very much under the radar of industry and government. As new small enterprises they face many expected barriers - funding, backhaul costs, technical deployment etc. Many cited digital inclusion as an important driver for their projects.

Q: Digital inclusion - how do people achieve this? A: Not really known.

Q: Could we get more regulation from OFCOM by disclosing this?

Balance to be struck; make info public, will it stimulate funding or conflict/regulation? Can flood each other's networks [with hot antennas], not traceable. So a big role for regulation. Manage noise levels.

A blizzard of OFCOM docs, but most on BT; some on spectrum regulation.

One argument is using unlicensed spectrum. Regulator not interested because ISM [industrial, scientific, and medical band. Argument is that CBNs are offering telecom services; huge layer of cost if so. You look like a telco, smell like a telco.

BT has more lawyers than OFCOM has staff. Many OFCOM staff are ex-BT. Just don't see it or aren't interested. BT regulations are terrifying; pleaded ignorance and ignore it so far.

Have user logs for last 5 years; what if police ask for it? Under the law you have to give them. Big ISPs asked to keep these logs. Communication Service Providers Act asks for a huge list of bits of data to be retained.

Q: So what to do with this data?

Don't suggest we keep this secret, but need to be ready with lobbying and responses. In the political sphere there's a warm fuzzy feeling about CBNs. Turning this into action and cash is harder work. Lots of it, but doesn't seem big - only 12000 people!?! If seen as a telco, get regulated. Part of inclusion, regenertion debate, get funded. OFCOM not part of that latter debate.

Large players are being subsidised; we aren't. Get a soft govt contract to build infrastructure, might not be corrupt, but isn't "straight" economically. Subsidy may come through regeneration, cohesion etc. So there's an offensive argument to play, don't just need to be defensive. Government idea of action is to run conferences. There is some loan money, but not designed for these uses.

Most CBNs beyond range of current ISPs, so they don't care. But after that we're in direct competition. Many are urban, too. London, Bradford, etc. One of largest nets is in Manchester.

Some CBNs see re-sale of ADSL as being OK. ADSL useful for VoIP backhaul compared to satellite. May lose a few subs, CBNs allow ADSL to be shared out.

BT - can afford to give away modems. But CBNs have lack of access to capital. Can recoup over time. Forces to charge for installation. People are willing to pay to have someone come round to install DSL etc and clean up computer. People happy to pay for water plumber, but what about IT plumber? Should CBNs try to compete directly with BT (e.g. same sales and pricing policies)?

Some not interested in low cost per se, but interested in more bandwidth to do interesting things. So you might thing GBP5-10 a month for low-income households is the thing; but could be new apps like videocast withing the community. Need to set out to highlight differentiators other than price.

Lots of anti-BT feeling among the users. Prefer local support, small org support. More than half have waited for BT's ADSL rather than go for CBN. Most of original motivation was because BT wasn't getting ADSL rolled out. ADSL exchange, but lines often old and poor quality, may not be full coverage. Example of big rural UK estate, want to offer connectivity everywhere. CBNs much more flexible in terms of technology. Can do channel bonding quickly, stay ahead of BT.

Easynet want to launch 18Mbit service, can't because of interference between pairs.

Low-income social inclusion -- difficult, because they need more support, less IT literate. Opens network to more risks. Value BB less than people who use it all the time. May be fickle, enticed by big operators.

Head of govt DTI, 2008 universal BB decree from govt, unsure what it means. How can govt create connectivity in a situation with 3m poorest people on tax credits?

Aren't many people in BB world that can meet Cybermoor's [a UK CBN] penetration.

Not just an economic issue. Costs are comparable to other forms of entertainment they pay for. Just isn't a priority for these people. Need to be relevant services that make digital inclusion a desirable objective. No point in having a connection if can't afford direct charges for content & service. BBC's project for archives online. CC license. Can re-shape it. People pay for Sky, can offer alternative to Sky of streamed sports broadcasts.

A difference between inclusion and enclosure. Might not want to be there. Have cynicism around reasons for digital inclusion. Social workers, teachers and polititians are laggards in tech use; end users are often ahead of them. But these are people who affect access to connectivity for those at the bottom end of the economic stack. Remote monitoring of elderly people -- can stay in their homes. Huge community benefit. Don't have to move to some old people's home 50-60 miles away. Village has 100 homes, over 60 members. A wonderful community effort. Finished 15 bottles of wine between 22 people who came to last meeting! Parish council minutes go around by email. Lots of fun. Ran a course, co-op gave a grant.

Power rating in rural areas? Less of an issue than people say, can use different bands (e.g. 5.8GHz), directional antennas, more dense meshes. Mobile phone coverage out-of-range in many CBN areas.

VoIP: many CBNs interested. Generate another form of revenue, an add-on service. Have spoken several times. Approach - make it work, then worry about the regulators. Unlikely to get closed down. Need to be mutual to avoid public service regulations. Need everyone to be a memeber. VoIP - who is actually providing the service -- breakout can be in another country. VoIP PBX is just an app server, beyond regulation. Voice charges are very low, not a revenue source. BT international happy with VoIP - generates traffic; naional side doesn't want it.

Q: Ts&Cs as a differentiator? A: Not understood by public. Community media services are differentiator. BB content only seen as exension of existing paradigm -- centralised production and distribution. Networking transmitting more than it receives. (Not just viruses.) Non-asymmetiric a selling point. ADSL avoid ruining the hosting services. 256kbit up not suitable for multiple home workers. Not willing to go from GBP80 to 1000. Examples of small businesses making video ads, sales.

Village web site can regenerate socialisation. Can up load own pages. How to make village shop go live? Send them your shopping list? Local e-commerce applications. Have services for receiving orders to local village shop, organising deliveries. [CBN is embedded in wider community efforts.] Village shop would need access to library of photos of products? Most people know what village shop sells, know the goods. People's orders very predictable.

Cybermoor - Daniel Heery

www.cybermoor.org

Lots of videos, forums (caused some controversy). Creative commons/Media Trust license - hybrid. Video which didn't have permission to show a kid. Amazon associate, brings in some money. Income-based charges. Just not worth doing VoD for Britney - you only get 2% of the revenue. Better to originate your own stuff.

Work on a project basis; lookin at community radio. Local bands, produce their music. Difficult to get people to pay to subscribe to local TV. Will pay 20 quid/month for Sky Movies, but nothing for local stuff. Have local radio on IP. Some content not routed to public Internet.

Should CBNs be putting their effort into these non-connectivity issues?

Oxfordshire Rural Broadband

Circles can be quite incestous, hard to draw new people in. People who already have BT broadband; draw them in. Healthcheck process (chamber of commerce, council, housing agency, etc.) gives SWOT on local area. List of projects. Looking at how technology can be applied to make these easier and cheaper. Community BB is a backdoor benefit of some of these. One village trying to put together a museum. Too expensive to get a building. Farmer has a barn full of exhbits. Doing a study to use a mix of mesh and RFID, and exhibits distributed around the community. Tourists then wander around

CCTV: GBP250k for dark fiber alone to police station, 30k for community network.

Have abandoned commercial hosting, do own hosting. VoD, open source. Have 20 nodes in network. Small area, densely packed. Need adaptive power system.

Cameroon community in east London making content, which is then pirated by the commercial operators in Cameroon!

Access vs. service provision; Internet was a solution to an existing problem. VoD is unusual in that users create channels and compete to do the scheduling.

Canada - TiVo on steroids; commercial satellite recorded and turned into VoD. Legal, has a CATV license. Timeslip TV. Don't worry about disk space, what to pull down. Cheap when done for everyone. In UK, can't get Channel 5, Freeview in places. Could use this instead.

Can have local school acting as a mesh server (not Internet uplink). School digital content available to the community.

School where kids can go home and re-run all the lessons. Failing school (Kingswood) adopted multimedia and much improved. Whole teaching methods changed; teachers re-trained.

Brian Condon - End Game for the Telcos?

Dilemmas:

Regulation as telco vs. as a economic and social development agent

Offensive strategy to generate support vs reproduce telco modes of behavior

Choice and flexibility vs safety of big brand

Match local needs vs sub-and-spoke consumer/producer

Access provision vs. service provision

Publisher filter/gateway vs. highly distributed control

When does a LAN become a WAN and get regulated?

Community network roaming?

Single source of purchase for longhaul, do a deal with independent community fiber provider?

Get BBC etc involved for political support?

Roaming - Ts&Cs for roaming, traceability, wouldn't open network until can protect node owners.

Options valued

I can commend this post by Ben Hyde on the option value of PCI expansion slots in PCs. In whose interests are all those unused slots deployed? It reminds me of this paper [MS Word] on option value in networking.

As David Isenberg once reminded me:

Option value increases under uncertainty -- if the future is relatively certain, then own the stock, but if uncertain, own the option.

A stupid network has high option value. But it lacks an equivalent of Microsoft or Intel who desire the rest of the value chain to be commoditised; the core optical and wireless transmission technologies aren't easily monopolised. Players like Qualcomm aren't nearly powerful enough to dictate the use of their network technologies. (It also partly explains why they fear Intel so much. Intel understands the game of profit-pool reallocation and has the deep pockets to play at the top table.) Hence stupid networks don't get powerful proponents, whereas stupid computers do.

We've seen a move from PCI cards to USB for peripherals. At the same time there's been an explosion of choice in consumer electronics. Ten years ago, your Gizmodo would have been a monthly, not hourly, publication. The shift to USB reflects the high option value of an open architecture, and the premium available for placing that option value in a more convenient form.

The move to representing documents as XML is a bit like the use of IP networks. Both are "inefficient" from a static viewpoint. VoIP is less efficient (from a bandwidth, latency and jiter perspective) than a dedicated network. XML is inefficient in CPU and storage compared to a purpose-specific file format. But XML maximises option value; you can transmit and transform data into places and forms not forseen when the data was first stored. Likewise, Oracle became dominant because its storage technology (relational) had higher option value than the competing (but faster and more "efficient") hierarchical databases of the time. HTTP allowed headers to be added at will.

Virginia Postrel famously wrote in The Future and Its Enemies that the world boils down to "statists" and "dynamicists", who respectively repel and embrace change. Mammals are also dynamicists; all that warm blood is "inefficiently hot", but the option value of a wider habitat makes up for it. SQL, XML, PCI, USB, HTTP and IP are all dynamicist ("stupid") technologies.

Dynamicists tend to outpace statists because they're more adaptable in a raw Darwinian sense. Yet how often do we consider option value when creating standards and technologies? How often do we lock ourselves into the known requirements rather than embrace the unknown ones?

Product microreview: BT Communicator

What it is: BT's PC VoIP client that comes bundled with the UK version of Yahoo! Messgenger. Calls get billed to your home phone. Supposedly takes incoming calls, buy I rarely get those on my landline unless it's telemarketing crap.

What it looks like:

What I like about it:

• Good sound quality.
• No need to worry about pre-paid balances. Just keep talking.
• Ability to call numbers normally only reachable from inside UK wherever I am.
• Integrates with Yahoo address book.
• Can place tolled calls away from home without burdening host (parents, friends)
• Gives out home caller ID, with optional block (cool!)
• Clear itemised billing, easy to expense

What I dislike about it:

• Expensive call charges
• Can't cope with loss of Internet connection; have to re-start app
• Hopeless with NAT, UPnP firewalls
• Stupid log-in screen that asks if you have dial-up or broadband (just ping and guess, guys!)
• No integration with Yahoo! login
• Non-BT customers of Yahoo! UK pissed off by being forced to upgrade to version with BT communicator
• Zero incremental features beyond PSTN
• Can't get BT 1571 voicemail remotely (as if I was at home) without paying extra; but can't easily subscribe to the feature when away from home (needs BT account # which I can't remember 'cos it's in a drawer AT HOME)
• Can't even see if there's voicemail waiting
• Doesn't integrate with PC well -- no options on audio input, doesn't intelligently deal with things like speaker muted or microphone not plugged in, can't access Windows address book

Overall opinion

It's a useful program for extending the reach of your home phone, and good enough despite some rough edges. I use it for business calls away from home because I've experienced mixed results with SkypeOut. Hardly the wave of the future, but better than nothing.

Call me contrarian...

This might be bonkers, but I'm going to float it anyway.

Incumbent telcos should be banned from deploying fibre-optic cable access networks.

Yes, your read that right. Lay fiber to the home, get jail time. Before the men in white coats arrive, let me explain. The danger is that they'll use monopoly copper revenue to build a new access monopoly. You can either try to tackle this through regulation, or through structural separation. I personally favour that capital allocation for fibre networks be allocated totally independently of copper. Apart from anything, it will stimulate alternative ownership and funding approaches -- which would be too risky with an incumbent Sword of Damocles dangling by a fibre.

But isn't it unfair, you retort? Aren't you robbing the incumbent shareholders of the opportunities of technological progress?

My first reaction to this is to start yelling expletives and shaking my fists, but once I calm down I would suggest you look at the actual rate at which such technology has been deployed (not promised), and how long such technology has been available. You aren't going to be robbing the shareholders of much, believe me.

So the incumbents become permanent custodians of the copper network. Just deregulate the whole thing. Rip all the pages out of the rulebook. Give their shareholders a little cheer for a year or two. And then let the cash gusher encourage new access entrants, safe in the knowledge they have nothing to fear apart from real competition.

Ooh, that's the doorbell. I'd better go now.

Overphishing concerns

Just to follow-up on some interesting stuff on James Seng's blog on spoofing and phishing attacks.

Take a look at this article on how Netscape 8 includes anti-phishing techniques. Notably, there's a "traffic light" icon to say whether a site is known to be kosher, a fraud, or unclassified.

A weakness of cell phones is that they lack a second user interface channel through which to express trust. Although there's normally a "reserved area" of the screen for the signal strength, battery, etc., this may be imitated by Java or Symbian apps that take over the whole screen. It's a weak security mechanism. Anyone can draw a convincing padlock. Various types of clever mutual-authentication systems exist, but require user education and mass adoption. Ultimately, metadata on trust is best sent via a different channel than the message whose authenticity may be in doubt.

So why not create a second communication channel on the phone to convey trust information? My favourite approach is to re-invent the logo. Every phone comes with a printed manufacturer or operator logo. This should be an LED, multi-coloured if necessary. It's the equivalent of Netscape's traffic light, and is a powerful associator of the trust message with the network or handset vendor's brand. (Of course, a true end-to-end fanatic would want anyone on the net be able to be the trust provider.) It's not enough on its own to combat spoofing, but a winking red LED could do a lot to tell you this isn't really your bank asking you for your PIN.

A problem of *inter*networking is that trust isn't transitive; you'll never rid the Internet of spoofed packets and misrepresented identity. The natural defence of the integrated monopoly network provider is to appeal to safety and security; it's the achilles heel of the Internet. Implausible promises of making the Net safer are likely to feature heavily in the plans of big telcos currying political favours.

Wiretrapped

Hang on a minute. Your ISP may want to charge you extra for hosting a server, or may want to deploy bandwidth shaping to throttle your P2P traffic. But who gave them permission to look beyond the IP headers?

Those TCP and UDP headers aren't needed to route the packet. Why isn't this an illegal wiretap? How on earth are Cisco et al getting away with selling deep packet inspection tools for public networks? Are terms of service that give your provider a blanket wiretap authority enforceable? How come?

Reader advice sought: is this actually legal in your jurisdiction? If so, how come? If not, why isn't the wiretap law enforced? How come people aren't up in arms about the privacy invasion?

(Note that intermediation services like Akamai's web caching are OK; although they "intercept" your packets, the packets have in fact reached their final destination. It just turns out that the IP address of, say, the Yahoo home page is a logical address, not a physical one. It's not a wiretap if the recipient authorises it.)

Laptops a-go-go

Hmmmm ... Negroponte sees 200m uber-cheap laptops for developing countries. Given a bit of cheap connectivity, I see 200m VoIP terminals. Sitty-talkies instead of walkie-talkies.

Hint to Negroponte: some commercial providers might pay to get, um, an icon on that desktop.

Good luck to him! The more smart nodes at the network edge, the better.

Towards a better spam filter

As far as I can tell, there's no prior art for this idea, so here goes.

One technique spammers use is to create odd spellings of Vgiara and P'ron to try to evade filters. They also stuff randomly generated text at the end of mails with qskyrjsna words.

I'd like to propose a modest enhancement to spam filters to trap these. I call it a Bayesian Trigraph Filter.

Traditional Bayesian filters use the frequencies of whole words to score mail as spam. The word "prescription" might be 75% likely to indicate spam; "baccarat" is 99%; and "disintermediation", 0%. Bayes's algorithm just says how to accumulate and combine these. Words not seen by the filter before are assumed to be mildly spammy. The problem with this is that "new" words are often legitimate ones, either of a technical nature, or just text quoted in a foreign language.

My additional filtering stage is to break up the text into sequential trigaphs. A trigraph is a simply a sequence of three letters. An example:

Input: [ "The quick brown fox." ]

Output: [ "The", "he ", "e q", " qu", "qui", "uic", "ick", "ck ", "k b", " br", "bro", "row", "own", "wn ", "n f", " fo", "fox", "ox." ]

Now treat the trigraph sequence as the words of the input document, and apply standard Bayesian filtering. Certain letter combinations ("oth") are common; others are highly unlikely ("qqh"). Lots of mis-spellings generate unlikely combinations. Deliberate made-up "salt" words to overload your spam dictionary are also likely to be rejected.

Advantages: fast, easy, simple, unlikely to generate significant false positives.

Disadvantages: Might trip up on embedded uuencoded data, which is the one case where you deliberately have strange letter combinations; doesn't do anything for spams padded with valid corpus text.

Letter trigraph frequencies will vary between languages, but the filter will quickly learn which languages you converse in without having to learn the entire repertoire of words in those languages.

This approach would also be good for providing an initial seed for an untrained spam filter, when the user-specific Bayes dictionary is empty. Your trigraph distribution is much more likely to be similar to other people than your word use.

I'm due to upgrade my personal spam filter (dspam) at some point soon, so I may whip together some procmail/formail scripts and see how it goes. I've got a pretty decent personal corpus to test against. I'll let you know how it goes!

A taxing thought

I've seen quite a lot of discussion recently about proposed new taxation of connectivity.

This is an unavoidable outcome of the application layer becoming disconnected from the connectivity layer. Regulation at the application level will die because it creates an economic disincentive to itself. The applications just move to another jurisdiction.

So we'll see communications taxation re-emerge as a tax on Internet access and ISP service, until such time as mesh networking takes off and people arrange their own network connectivity.

The missing wildcard is identity. As I've mentioned before, a byproduct of a connectivity or service provision relationship is digital identity collateral. Without such collateral, the policing of abuse becomes hard.

How does this work? The provider can do harm to you if you do harm to others: they can tell the police where you live, who you are, what your bank account or credit card number is; or just simply cut off your device or service. "Free" connectivity (as in "beer" and "speech") doesn't provide this by default.

To date, you have to go to a centralised entity to obtain the collateral-backed identity that makes interruptive services socially scale. I would therefore expect to see taxes move from voice service provision to the issuance of telephone numbers and even domain names. Why? Because of the difficulty of building decentralised identity systems that enbed identity collateral.

The whole debate about "taxing VoIP" is a misnomer, because IP unbundles the elements of PSTN voice service. The real debate is about what those bundled elements are (I'm at about 20 and counting!), and which (if any) should be taxed and regulated.

UPDATE: Check out Aswath's comment below. Also, how on earth do you tax access network connectivity once you have choices beyond dial-up? No really ... how do you do it? If it's a standard sales tax, no problem. But if you want to "supertax" connectivity, you have to define it. Does it only apply to "Internet access", to avoid tolling LANs? What factor to you vary your tax on -- just downstream bandwidth? Where does the "Internet" bit start if you put together a more ad-hoc network? How do you align jurisdiction with network architecture? It's a nightmare!

Keep it brief

I'm catching up with my news feeds, some quick shorts.

Disrupt from below

I notice that Uniden (of cordless phone fame) are getting into household VoIP.

If I were a big-name cellular handset manufacturer, I'd be worried. A Wi-Fi phone unencumbered by the history and price-discrimination of a carrier could start to sprout some pretty compelling features. Then disrupt upwards; people bring the phone to their vacation house, workplace and relatives. It's portable. Then a municipal or next-generation mobile wireless broadband network launches, and you take your home phone with you, not your old cell phone.

Flarion abandoned

Sprint-Nextel abandons Flarion trial. Predictable. If you're the guy in Lenexa, KS choosing a network technology, remember you have two options: A. EV-DO. Doesn't do VoIP well, reverse link doesn't scale well, laggy, needs an army of network engineers. B. FLASH-OFDM. More symmetric, low latency, ideal for VoIP. Perfect wireless stupid network. Few network engineers needed.

You choose A, every time.

Bundled telecom

So, BT have responded to OFCOM's challenge to design their own regulatory shackles. Unsurprisingly, they chose the, um, successful American model. Basically, local loop unbundling (LLU) is a dud. You need local loop transfer (LLT). Get it back in the hands of the users, who then contract someone to maintain and connect it for them. Might be individual, might be en-masse. All LLU does is keep pricing in the political arena, and BT's lawyers know how to play that game better than anyone. LLU is not free market competition, just a facsimile of it.