Yippee! Off on vacation for two weeks. No laptops allowed, by decree. Sorry for the slow rate of entries recently -- getting ready for a big journey with a baby is time-consuming.
We went to a party at the weekend. A PhD graduation celebration for one of my wife's friends. When you've been incarcerated with a ten-month old baby for, well, ten months, such social stimulation is most welcome. Unfortunately my conversation skills no longer extend beyond pointing at things and naming them. But no matter, a good time was had by all.
My wife's friend (from Finland, blonde, owns a Nokia phone, don't you love stereotypes) has two young cats, both from the same litter. Their pelts are very different colors, but when they stand together you can see the almost identical bone structure in their faces. They have the lean, focused look of hunters.
We learned a great trick at the party, one you can use to amaze friends and alarm neighbors at your own social gatherings. The trick requires three readily available ingredients. One. A laser pointer. Two. A carpeted floor, for plenty of claw traction. And three. A minimum of one cat.
The method is to shine the laser pointer in front of the cat. Cat pounces on laser spot. Spot moves rapidly back and forth. Cat moves rapidly back and forth. Spot moves in circles. Cat whizzes around like a spinning top. Cruel? Maybe. Fun? Absolutely.
Anyway, enough of that. In my inbox this morning was a link to the well-trailed news that Skype plan to offer paid-for interconnect to the PSTN (thanks, Robin). Now, Skype have also been getting into the hardware business, so this isn't the only string to their bow. But it highlights once again the fallacy that there is a future for recurring service revenue from vanilla VoIP. There isn't any.
For the hard of thinking, I'll recap the argument once again. The customer with a broadband connection has already paid postage and packing for the bitstream that carries the voice call. Increasingly, the public Internet is capable of moving those bits with acceptable latency and jitter to any other broadband-connected user anywhere else in the world. The result? High-quality full-duplex interactive audio, previously known as a phone call.
The only centrally mediated function is turning the persistent identifier of the callee into an IP address. That name could be a phone number, proprietary network identifier (like in Skype) or an IM/e-mail address. This look-up has a cost structure of close to zero. Anyone can create a competing namespace. And the phone number namespace is already regulated to prevent ownership by any one commercial entity. So no profit goldmine hiding under the mat there. Only a transient self-defeating market for VoIP to PSTN interconnect. The more VoIP there is, the less need for payment of tributes to centrally managed faux-circuit-switched service providers.
There's commercial value in creating new services, new customer hardware, and maybe even in the connectivity itself on a good day. There's also value in preventing unwanted connections. But not in recreating a dead business model of circuit voice.
So if you're looking to invest your own money in the VoIP boom, just be careful. You might be chasing prey that doesn't exist.
There's a corollary to the paradox of the best network.
The network becomes dumber and dumber, reducing opportunity to differentiate by embedded service. The customer's cost of Internet membership becomes lower and lower -- affordable to all -- and reduces the opportunity to engage in creative pricing of the services on that network. For wireless connectivity, the coverage reaches the "good enough" stage for the great majority of users.
This means that people are likely to have a choice of networks that meet their needs, with clear market signals as to where the best deal is, and products that are readily compared. Users are therefore likely to revise their choice on a frequent basis (until the municipal fiber reaches their front door, which is game over for everyone else). This makes the "first mile" connectivity market look a lot like the (horribly low-margin) long-distance voice market today. It's not unusual in long distance to have over 100% churn (i.e. if you had 1,000,000 customers on 1st January, you'll cumulatively lose over 1,000,000 customers during the year; you have to make that back up with new customers, some of whom will defect and need to be be replaced by yet more new customers if you want to stand still.)
The consequence? Telecom (meaning "bit haulage business") becomes less and less about networks. The network division of a telco has traditionally been the exalted realm of high priests of technology, beyond challenge or even control. The IT folks were seen as second-class citizens in the technology stakes. And the operational management, well -- they just don't know anything about anything, do they?
Instead of this, success in future is increasingly defined by non-network attributes. Your ability to cheaply acquire and rapidly provision customers; to accurately manage the credit you extend; pro-actively seek out and destroy the root causes of customer care costs; excel at billing and collections; juggle your finances better than anyone else; perfectly tailor handset, spectrum and network supply chain management and inventory to demand; ramrod a few bundled third party communication services onto the bill; and so on. Anything to snatch and hold on to uppity churnaphilic customers.
That's quite a change in the internal political landscape in telcoland.
The recent Harvard Business Review article on operational innovation offers a clue to how to survive as a telco in the dumb pipe world. Get really good at operations. At the right time, sell your network off and become a service company.
We're already seeing horizontal layering like this today with networks assembled from multiple affiliates (informed readers will know who I'm thinking of...). The affiliates do the local network build-out, the central telco supplies the brand and enabling services. Another example is Wayport, a lonely survivor from the WiFi bubble, which aggregates the provisioning and billing across multiple networks. (The demise of Cometa just adds more fuel to the argument that network operation is a dunce's business.)
Perhaps we're heading towards a future where instead of a dozen large vertically-organized telcos in the US, you'll have thousands of niche marketing-based companies. These are the front-end to the customer, and may be embedded in other businesses that have an existing customer relationship to manage. (How long until Wal-Mart enters telephony on its own terms as a service provider rather than reselling phone cards?) You'll then have a few dozen platform companies that provide the enabling services. These will either excel at process innovation or have control a few key platform technologies. Think of Amdocs and Convergys; Microsoft and Sun; systems integrators and their outsourcing arms; and spun-out chunks of telco that excel at specific process steps. Finally there's a handful of long-haul network operators gently revolving through the bankruptcy cleansing cycle.
(Incidentally, the moves of IBM into the operational part of the outsourcing business -- they don't just built your IT infrastructure, they'll also operate your business for you -- suggests that they want to own the process innovation. They can see a way to suck the lifeblood out of their own outsourcing customers!)
Of course, you might try to sail against the tide, and focus on things you can control. Stiffly try to preserve the vertical structure. Deploy locked-down set-top boxes and handsets that only can operate on your network on your terms. Set up thin pipes with connectivity rationed through quality of disservice measures. Distribute copyright-controlled media content and DRM, rather than slippery user-generated (and user-controlled) messages. As a tactic to generate cash, that's fine and dandy. As a strategy to survive, it's suicide.
UPDATE: More on structural separation of telcos here. A distinction not really drawn is the difference between application service and customer service. Should the application platform (those SS7 switches, HTTP proxies, SMS gateways, etc.) be divested together with the connectivity or separately?
Worth a read is George Gilder's testimony to the US Senate (via /.). In a nutshell, he argues that horizontal regulation of the communications industry (into open access pipes, network services and applications) is unnecessary and counter-productive. It precludes useful (indeed inevitable) localized vertical integration of the industry. In George's own words:
Layering proponents, however, make a fundamental error. They ignore ever changing trade-offs between integration and modularization that are among the most profound and strategic decisions any company in any industry makes. They disavow Harvard Business professor Clayton Christensen's theorems that dictate when modularization, or "layering," is advisable, and when integration is far more likely to yield success. For example, the separation of content and conduit—the notion that bandwidth providers should focus on delivering robust, high-speed connections while allowing hundreds of millions of professionals and amateurs to supply the content—is often a sound strategy. We have supported it from the beginning. But leading edge undershoot products (ones that are not yet good enough for the demands of the marketplace) like video-conferencing often require integration.
Such a horizonal re-regulation is being proposed by MCI. Personally, I too believe this is really, really, stupid. It's stupid at the semantic level, because they try to make false distinctions between applications and content. Is an ActiveX control in a web page content or code (cue another 10,000 lawyers)? It's stupid at the technical level, as Gilder points out, because sometimes there are good architectural reasons to integrate. And it's stupid at the economic level, because it presumes to define an outcome (horizontally layered industry) in preference to enforcing the full operation of the free market. Free markets are our way of experimenting, eliminating dumb ideas and identifying bright ones. If you don't like it, you'll be pleased to know that tourist visas to North Korea are now available. Or in George's more gentlemanly language...
Proponents of "layering", or "Net neutrality," or a free Internet "commons," assume there is one network, that it is sufficient and timeless, that no new networks are possible or needed. They want innovation on the edge, in the form of software apps and Wi-Fi attachments. Innovation in the core is either assumed or ignored. The logical conclusion, however, is that since the "best network"—the free commons—cannot make any money, there will be no network. And just how much innovation at the edge will there be if there is no innovation—no bandwidth—in the core?
Gilder's view is that end-to-end optically-switched fiber is the future for all fixed networks in most situations. The last word in dumb pipes is when the same photon you stuck in one end zips out the other, pristine and unmolested. The only value-add is an occasional gentle nudge in the general geographic direction of the desired end point.
Misguided regulation may be worse than no regulation. (As a thought experiment, what would have happened if Ma Bell hadn't been broken up? Where would the US be now?) Over to George...
The real threat to monopolize and paralyze the Internet is not the communications industry and its suppliers, but the premature modularizers and commoditizers, the proponents of the dream of some final government solution for the uncertainties of all life and commerce.
Fundamentally it's hard to truly gain market power in telecom when exposed to open competition. At the physical level, the transfer of electrical signals into photons and back again is a concept beyond ownership. Yes, you can patent specific ways of doing it. But unlike, say, Windows, the basic protocol came free with every copy of Universe 1.0. No matter how clever you are, you can't ultimately own the idea of popping a photon out of one fiber or radio receiver and stuffing a similar one down another output to get it closer to its destination. Most of the protocols at the next few layers -- IP, BGP, MPLS, TCP, Ethernet, etc. -- are de facto public domain.
There are only two ways to gain true market power in bit haulage. One is to come up with an innovative new (patented, proprietary) physical media access technology like CDMA, FLASH-OFDM or DWDM. The other is to legislate away competition.
As I've argued previously, what telecom needs is less regulation, not more. And what regulation that there is needs to be focused on rigorous enforcement of competition policy. To the point where you might spice the mix with some pro-active rules to encourage new entrants and discourage predatory pricing. The victims of dominant incumbents wielding below-cost pricing are not likely to be in much of a position to complain when they're dead or dying. Justice delayed, etc.
Notwithstanding the above, Gilder has garnered himself a reputation of being the false prophet of false profit, as this Slashdot comment shows:
George Gilder seems to have succeeded solely on the basis of his belief in his own power as a prophet of the future. As those who subscribed to his stock market newsletter found, he was a legend in his own mind, not in reality.
[...] So Gilder reinvented himself as a technology guru. The fact that he has no background what-so-ever in science or technology did not stop him. He interviewed those who did and wrote up his impressions in breathless terms.
The peak of Gilder's trajectory was his stock market newsletter which had thousands of subscribers who were willing to pay thousands of dollars for the privilege of reading the thoughts of the master. [...]
Then there was the fall. As the 2000 stock market crash erased the value of many of the stocks that Gilder touted, his subscribers deserted him in droves, much poorer for the experience. Gilder had invested in the stocks that he hyped and his investments were largely wiped out. Gilder was also making money holding conferences and was left with conference committements and no attendees. In the end he was heavily in debt, his bubble wealth wiped out.
Well, any pundit willing to put his money where his mouth is gets bonus whuffie in my book. It's far riskier placing bets on individual players and specific timing of events than on general trends. For instance, the general trend has been a great increase in human longevity over the past century. But you'd be hard pressed to put that fact to use standing in downtown Baghdad playing whack-a-mob, where life expectancy has taken a nasty turn for the worse. However, such knowledge could be useful in deciding whether to invest in a portfolio of retirement homes or the music industry. Strategic insight doesn't necessarily translate directly into tactical brilliance.
By coincidence I recently finished reading Gilder's Telecosm. It's a good book, although I suspect George got a bit overexcited buying value packs of Hyperbole at Costco. Once or twice you wish he'd save some face by simply stating "I didn't do so well at physics in school", and drop a page or two of literary arm-waving. But the central thesis holds. We're moving from connectivity malnutrition to gluttonous data obesity; fiber and wirless broadband are the drivers; and there's a lot of money to be made and lost in the change. To save you from the effort of reading (or thinking too hard), here's some juicy quotes.
Firstly, the physics of an industry ripple all the way up to influence its macroeconomic structure.
Explained Chris Harder, cief of R&D at Uniphase in Switzerland, "In semiconductors, you get more and more transistors on one chip and everything improves. But in telecom, you don't necessarily want more lasers on one chip. You want more light pulses emitted into the fiber. integration may not help at all."
In this view, photonics and electronics are inorexably different. At the heart of integrated circuitry is the phenomenon of voltage fanout -- a voltage divider. You can divide a voltage many times without reducing it. Thus millions of transistors can be activated across the surface of a chip without attenuating the voltage. But divide photonic energy and you attenuate it. Each time you split it you have less. Thus integration is more difficult.
Unfortunately, George immediately puffs JDS as being the future Intel of telecom, without getting it himself: there fundamentally isn't an integration interface that can be "owned" in the same way as in electonics. There never will be an "Intel of telecom".
Next, on photonics vs. electronics:
Traffic on cable coax runs at radio frequencies. When it moves into glowing silica threads, it leaps upward in the spectrum. The lasers of fiber optics emit waves that cycle a million times faster than do oscillators in a cable TV system. [...] If [the Internet] were a rainbow, the center of intensity would move up [over time] from red through green toward violet. If it were a meteor, the Doppler blue shift of the Internet would suggest it is approaching you.
The change in "colors" signals a shift in the nature of the network. Rising toward the terahertz region, colors become more and more difficult to manage electronically. So as the core of the net becomes brighter, it also necessarily becomes dumber.
The take-away here is that "dumb pipe" is not "IP pipe". The Internet (as currently physically embodied) is too complex and electronic. [I won't comment further on the confusion between frequency and magnitude in George's text. He's a much-needed storyteller, not a scientist.]
On information theory:
In order for a message to be high entropy (full of information) the carrier must be low entropy (empty of information). In the ideal system, the complexity is in the message, not the medium. By eliminating the entropy from networks--moving them onto the pure waves of light--you can increase their ability to bear information. Another word for a low entropy carrier is a dumb network. The dumber the network the more intelligence it can carry.
Amen, brother.
George also has some useful insights in his afterword into peer-to-peer networking, and recaps the "storewidth" arguments of Negroponte. But I think you ought to do George the favor of reading his book if you want to find out more. Don't you agree?
UPDATE: More George here
So, smaller US carriers are doing an AT&T Wireless and failing to prepare for number portability in the hope that the FCC will lose nerve and delay the whole thing.
The answer to all this is simple to describe, if complex to implement. Stop leasing phone number to phone companies. Sell them direct to the public. As fungible property. Just like domain names. Want to use a phone number for your VoIP system? Just buy one. Need a virtual fax service? Buy another.
Number portability would go away. I would own the number. I would control who it gets delegated to. Now my telco might choose to also bundle number registration with voice service, but nothing would force me to buy an identity from them.
It would also have a liberating effect on innovation, and indeed a rejuvenating effect on telcos. Imagine that any two guys in a garage can create an innovative communications service which uses telephone numbers. No need to become a registered telecommunications provider. No need to sub-let phone numbers from helpful local carriers. Just do it.
The phone number system is in competition with other namespaces. (Although the telcos can't feel it, and don't realize they're losing ground.) If you're a telco, you want new lifeblood injected into the current phone number namespace. For the next generation of customers, telephony is a minor feature of an instant messaging client. Need I say more?
Barriers to execution? Well, number exhaust has been one barrier to increased issuance of telephone numbers. But really, who cares? If there is a genuine shortage, why not let a market pricing mechanism decide who values them?
Also, the tying of tariffs to area codes (local vs. regional vs. long-distance calling) makes some numbers more valuable than others. But "area code arbitrage" is already a reality with VoIP. My Vonage code is in Seattle, my home in Kansas. Friends in Seattle call for free. Ones in Kansas don't. The system's obsolete. Charging by distance for narrowband voice is about as logical as charging by the number of digits in the phone number. It doesn't make any sense. The price no longer has any relation to value or service cost.
Doesn't ENUM do all this for us? Well, yes and no. No, because it's just a technology. No, because it doesn't change the governance of phone number issuance (and that's a country-specific thing anyway.) And yes, it could play a part if done right.
So, to recap. Number portability is a hideous con. The public has paid a fortune to perpetuate a governance model that's counter to their own interests. You don't expect the government to name your children, so why accept telco control of your virtual identity? Why lamely accept the absence of property rights? Isn't it time to end this telephonic serfdom?
I should be in bed, tomorrow is a long day. We're off to Las Vegas to meet up with the inlaws who are on their road trip. A weekend in Death Valley is in prospect. But I have a burning idea I just have to get out there, right now.
I was re-reading this Fortune article by David Isenberg on how DSL and cable have perverse economics and fiber is the glorious future. My mind started to dream about panglossian fiber sprouting from every wall outlet in every home and office. And then it hit me.
A potato has no value on its own. You have to cook it.
He really does need some sleep, you're thinking. Here's the idea. Consider networking in abstract. Networks enable connections. Faster networks enable more connections. More pervasive networks enable more connections. But a physical connection alone has no value. You have to run a service over it. A connection on its own is a potato. You can't eat it raw.
But not all application-layer connections create value. At any one time only a finite number of services have been invented. Only a modest number of end points create value to you in relation to that service. There are only so many people you want to talk to, trade with, get laid with [maybe this theory isn't so hot after all...] A connection from any other end point may have zero or negative value. (Think: e-mail spam, being stalked on a dating site, hassled on a social networking service to make introductions for remote strangers.) And those value-less connections are essentially without bound in number and frequency. They can easily swamp the positive value of the system.
The paradox of the paradox of the best network is that increased connectivity alone does not guarantee any increase in end-user value. Indeed, it may destroy it. For example, the spread of big-pipe broadband in east Asia has enabled a torrent of spam and denial of service attacks against the thin and brittle pipes into US and European homes. The entirety of the Internet infrastructure in Russia has net negative value to me personally. Three of the banned comment spam IP addresses on this site are Russian. I get hardly any genuine visits from Russia. My parents might be better off if their ISP refused all connection from Chinese hosts! There's nobody in China they feel an urge to converse with. They don't run a telecom website with a global audience. They just want to chat with their family and friends, and shop for a retirement home. And that's it. Reject everything from Xianggang and they're better off.
Just as computer-mediated social networks have poor social scaling attributes, pervasive unmediated data networks may have negative value scaling attributes. Wall to wall fiber isn't good enough. Connectivity isn't value in the eyes of the user. In can only be used as an ingredient to something else. You don't get the best value by eating all the food at the all-you-can-eat buffet.
Now, when you create a network, you do it to span a geography. Your personal influence -- your land, the sites you control, the people you know -- is very limited. At some point you stray outside your domain of control. At that point you need to work with a stranger who can get your lump of data nearer to its physical destination.
Traditionally, you pay a network service provider to do this on your behalf. Alternatives you can envision in the future might include mesh networks, municipal networks, or free community networks. Even if no payment is made, you're still expected to do something in return. It might only be to patronize a city district and buy a coffee. In all cases, there is what our lawyer friends call consideration. Namely,
".. some right, interest, profit or benefit accruing to the one party or some forbearance, detriment, loss or responsibility given, suffered or undertaken by the other."
So there is a chain of reciprocacy of mutual obligation embedded in the Internet. It could be to make money flow in the opposite direction to the packets. But it doesn't have to be cash. It could be an agreement to route some packets in the opposite direction. Supply copious ale when you next meet. Whatever.
The trick to going beyond the fibersphere is to understand and utilize this thin relationship network to help mitigate the negative-value connections. What can those around you state about your identity and connection-worthiness? If the people who are so willing to pass on your packet aren't willing to say anyting about you, why should I take the risk your incoming connection will have negative value to me?
Funninly enough, my degree was in Mathematics and Computation. I spent years studing formal methods of proving the correctness of a program in relation to a propositional calculus specification. But in the end I decided the thing was a waste of time, because the most important variable in the equation -- the human being coming up with the requirements that composed the specification -- wasn't appearing anywhere in the model.
Likewise, end-to-endism is stuck in a world of routing and connectivity. It is blind to value. There is no philosophical system behind it as to how value is created for people. Ubiquitous fiber and wireless ultrabroadband isn't the target end state. We need more. Giving people control of their connectivity isn't helpful if they don't have the information they need about the endpoints of the network to make informed decisions. The solution isn't "identity" or "reputation" or "trust". We probably don't have the vocabulary to describe it properly yet.
I have some ideas on what you can do to make that degenerate social network create value. I'll document them over the next few weeks. It might even help to rescue some knackered old telcos. But in the meantime, the future of telecom is clear. It looks like a spud that you bought from a stranger at the market, and devoured with gusto. But not raw.
Now I'm going to bed. Goodnight.
UPDATE: Or to put it all another way, now I've had some rest, the end-to-end argument optimizes a single connection (by assuming the least possible about the purpose of the connection). However, it doesn't necessarily optimize the value of the system to a user over multiple connections, not does it optimize the total system value over all users. The true goal is one of these latter two (a debate of its own), not a single connection.
It seems that Personal Mobile Gateway technology is slowly creeping towards the market (via Techdirt). These deivces enclose a wireless Internet backhaul connection and local wireless links (using Bluetooth, etc.) to gadgets that need the connectivity to the outside world. The result is a Personal Area Network spanning your bald patch to your ingrown toenails.
What's interesting about these is that they physcially separate service and connectivity into different devices. This is likely to cause heartburn to wireless network operators, who delight in controlling the devices on their network and extracting maximum revenue from service price discrimination. Even worse (from their perspective) is that these devices are likely to create an ecosystem of application devices that is totally outside carrier control. Once you have a personal gateway on your belt clip, you're not going to take kindly to your carrier telling you which gizmos you can carry on your person.
My take is that the early adopter market is barely ready for these devices. The obvious users will be in vertical enterprise applications. This is nowhere near ready as a consumer play. Cellular data is too expensive as backhaul. WiFi has too short a range for pervasive coverage of a factory or office building. Alternatives are either highly proprietary or pie-in-the-sky. Maybe give it another two years. But this technology is inevitable and is yet another nail in the coffin of the network operator having control over the application services.
What I'm about to say seems obvious, but only in retrospect. You'll already know it all, just won't have said it out loud.
The voice calling industry has been shifting from landline to cellular for a number of years. Cellular minutes of use have exploded while landline talk has stagnated. No news there. [This is bleedin' obvious already. Ed.]
Landline networks let you place calls from one fixed point in space to any other point on the network. One to many. Mobile calls can be made from any point in space (within coverage) to any other point on the network. Many to many. [You're fired. Ed.]
This difference is the driver that creates fundamentally different economic structures for fixed and mobile telephony. It isn't the wirelessness, but the mobility that matters. Why?
Well, when you have a fixed network you can use any technology you like to do the "local loop". Cable, MMDS, laser, copper twisted pair, glass fiber, etc. You just have the appropriate adapter box in your home or office to let you plug into that physical layer technology.
From your perspective, you don't care at all what the technology is or who supplies it. From your network operator's perspective, they want a vendor with low prices likely to stay in business for a while. A choice of interchangeable vendors is a bonus. But a niche technology is OK too.
This implies a highly fragmented market for fixed access technologies and network operators. The exception is for those serving multinational companies who want seamless global private networks. Then size matters.
Mobile or portable networks (including WiFi) don't work this way. When you move around, you need to interface with whatever the "loop" technology is in your physical vicinity. You need compatibility. That makes for a co-ordination problem. That may be solved by a single company building an ocean-to-ocean network. Or by a suite of companies where billing, access provisioning and network operation may be separate functions (as with WiFi). Either way, they are going to charge some form of economic rent for the privilege or performing that "works anywhere" co-ordination function.
Dampening this down, however, is that to make the co-ordination function work at all you need technical standards, which in turn encourage new entrants and open markets. So the mobile wireless network industry is likely to be consolidated with a few large moderately profitable network operators.
This is why companies like Qualcomm that can corner a mobility technology make buckets of money, when network operators don't. It's why Nokia is profitable: they own their own little compatibility ecosystem. And it's why I'd invest in a company that has a unique 802.20-compatible technology (built for mobility from day 1) over an 802.16 WiMax technology (kludged for mobility from day 100).
The consequence of the above is that the future for network operators is at one of two extremes. Firstly, very very large. This could possibly be integrated across wireless, national and intercontinental backhaul and local fixed delivery. The scale for fixed networks can be leveraged through advantageous peering deals, probably driven by owning key backbone fiber.
Very large mobile network operators will dominate, because smaller competitors will be driven out by steep roaming and termination charges. But they won't be particularly profitable because the compatibility will be owned by technology companies, handset manufacturers, operating systems vendors and provisioning aggregators. (Interestingly, this is a bit like the global airline industry consolidating into alliances. But unlike telecom, there's no roaming or termination charge issue with stepping off an EasyJet plane onto a British Airways one. Thus smaller plays stay in the game. "I'm sorry Sir, you appear to have travelled here on a discount operator's flight. We're going to have to surcharge your ticket.")
The other extreme is very small, possibly even built by the users themselves (think: WiFi, packet radio ham networks). The less mobile the requirement, the smaller the size of the market sub-division. So if you're in the fixed wireless broadband business, you want to follow a strategic path of customer intimacy and product differentiation to fit a niche, not scale or low price.
And everything in the middle will be eaten of will fade away. Sort of like Goldilocks, but with only hot and cold porridge on offer. No fairy-tale ending for telecom!
So, to wrap up, the structure of the communications industry isn't an accident -- it's the inevitable consequence of mixing economics and physics. The invisible hand meets the unseen wave.
Noticed the mad rush of all the traditional telcos to bundle wireless, local, long distance and broadband? If not, you need to get in more and waste more time online and watching TV adverts. It's a big deal.
The theory of bundling is well known. Aggregate a bunch of things together where different people tend to place widely different values on them. Setting a price for the component parts results in under-charging some people while others walk away from the deal. Price them together and you smooth this out. It helps you get the best overall revenue stream. As a bonus, those bundled customers tend to have much lower churn.
The bundling approach of current telcos only being done because the strategic options are so limited.
A product differentiation strategy is tough because you're hung by your own petard of the PSTN: meaningful service innovation for the core voice product is nearly impossible. Wireless coverage is at the "good enough" stage for most network operators. (Many of the differences are perceived, not real. Discount your own anecdotes.)
Low price is painful as it's hard to build in low cost. You pay more-or-less the same as everyone else for truck rolls, spectrum and switch gear. There's only so much you can pare back on marketing. There doesn't seem to be a cost innovation in the offing equivalent to Wal-Mart's cross-docking, or Dell's direct sales and supply chain integration model.
And customer intimacy only goes so far. As long as you bill accurately and the service doesn't go titsup too often, nobody really cares that much. (On the other hand, screw it up and you'll be joining AT&T Wireless as a case study in business schools everywhere.)
Even worse, the research it is based on is misleading. Ask your customers if they would like to eliminate the hassle of multiple bills, passwords, websites and support numbers, and of course they say a big Yes. But they're never shown the real choices they will face in future. Would you dump local and long distance service entirely if you could get all the voice you needed for $15 from Vonage? $10? $5? Would you say the same thing if someone offered a cellphone which integrated beautifully with your Playstation? Your Mac? Your Yahoo! service? Would you switch from DSL to wireless broadband if it became available in your area? To your shiny new municipal fiber system?
A nasty part of the one-bill approach is that the back office is so hard. To get the cost benefit you need to be able to unify your marketing, reporting, collections, activations, retail, channel partners, fraud management, network capacity planning, CRM, care, e-commerce, sales -- the list goes on and on. Apparently trivial stuff like matching customer records and verifying those matches turns out to be hard work. To make it work you need to update huge numbers of business processes and systems. Each implemented differently by each division, subsidiary or partner. It's like changing the wheels on a moving car.
Bundling also helps setup internal strategic conflict. Why would the marketing VP encourage people to cut the cord when your bonus payment depends on selling bundled service? This helps explain why companies like Vodafone are 100% focused on wireless. They believe that service innovation will outpace efforts to lock customers in via wireline service.
The revenue upside is also hard to execute on. As more and more "data" products are externally sourced and operated, you're dependent on being at the cutting edge of federated identity to make the user experience acceptable. Wireless operators may have a modern enough infrastructure to do this. Wireline networks generally don't -- and without them your bundle falls apart.
Microsoft and Yahoo! must be licking their lips at the opportunity to make their services "home base" on top of any and every dumb pipe.
So once everyone has the "one bill" approach we'll be back to square one. The price floor is set externally by the government and equipment vendors. Competition from fixed wireless broadband access will squeeze you hard. The political fence you built as protection becomes a prison as high quasi-regulatory charges incentivize switching. Customer service and intimacy runs out of steam pretty quick. And the demand for service innovation just piles up and up behind the PSTN dam. Eventually the pressure will be too great.
People want secure voice calls, presence, and IM integration. They want seamless interoperation with their other IP-based devices like PCs and Playstations. Elimination of barriers between their AOL address book and their voice communications services. They don't always know they want it. But once they experience it, there will be no going back.
The result of decades without any innovation for the PSTN could be a dam burst. A switchover to IP-based service could be shockingly quick, like the rise of dial-up access ten years ago, or fax ten years before that. And if all you've got is a defensive bundling strategy, you'll be left holding a bundle of nothing.
UPDATE: Tired of playing defense? Here's the offense strategy.
In Good to Great, Jim Collins proposes that one aspect that great companies get right is they have the right "denominator". In other words, they know which metric drives behaviour that aligns with their end strategic goal (rather than the nearest staging post on the way, or downstream effects like overall profitability).
Carrying forward my thoughts on energy use in telecom, perhaps we need a better measure to compare "dumb pipe" technologies. Telecom is one industry not upturned by the Death of Distance, because it itself is the industry tasked with masking out distance. The objective of a bit-haulage communications network is to cause the geographical migration of information. [Duh!] Disregarding Heisenberg and Einstein for a moment, all information is tied to a physical place at any instance in time. Every node in the network, be it optical or electrical, is there to direct the bits physically nearer to their destination.
So we want more miles. More bits. Fewer seconds to send those bits. And less energy to send them. Metres times bits as the numerator. Seconds times joules as the denominator. That makes bit-metres per Joule-second (bm/Js) our measure of the attractiveness of a communications technology. So unlike the usual method, we're not just looking at bits per second. A technology that can do 100Mb/sec but only projects those bits a metre is much worse than one that does 10Mb/sec to the other side of the world. Likewise if it takes a nuclear power station vs. an AAA battery to do it.
So let's take a few examples. Carrier pigeon with an aerodynamic flash drive strapped on its leg (we're not the first to try). First-class mail. FedExing hard drives around. A GSM phone. And a fiber-optic cable.
Carrier pigeon. 128Mb flash drive. That's 109 bits. Speed of pigeon 30mph. Weight of bird 500g. Metabolic rate 77Kcal/day = 77×4.2×103/3600/24 = 3.7kW (that's one hot bird!). Range 90 miles = 144,000m. 10,000 seconds. Result: (109x144×103)/(3.7×103x104x104) = 400bm/Js.
First-class mail plane. 2Kb of text on a letter. One letter 25g. One plane 50 tons of mail. Two million letters (we're packing them in!). 32 megabits of data. Boeing 757 fuel capacity 40,000 litres. Energy density of kerosene 46Mj/litre. Total energy 1.8TJ. Range 6000km. Cruising speed 900km/h. Flying time 24000s. Final result: (32×106x6×106)/(1.8×1012x24000) = 4×10-3 bm/Js.
Same plane. Filled with 200Gb hard drives, 1kg each inc. packaging. Assume we're limited by weight, not size. That's 50,000 drives, or 10,000,000Gb. About 1017 bits. Result of (1017x6×106)/(1.8×1012x24000) = 1.4×109 bm/Js.
Cell phone. Random Nokia model. Battery capacity 1300mAh. Duration 4h = 14,400s. Energy expended 1.3Ahx3Vx3600s = 14kJ. Data rate (reality, not spec. sheet): 20Kbps. Bits sent = 20,000 × 4h = 288mbit. Distance to tower: 10,000m. Result: 288×106x104/(14×103x4×3600) = 14,200 bm/Js.
Fiber optic: 5W laser. 1000gbit/sec. 9,000,000m from USA to Japan. Result? 1012x9×106/5 =~ 2×1018 bm/Js.
A quick summary:
| Method | Approx log10 bm/Jw |
| Snail mail | -3 |
| Carrier pigeon | 2 |
| Cell phone | 4 |
| Fedexed hard drives | 9 |
| Fiber optic | 18 |
Is this of any use to anyone? No idea, but it's one way to pass a Sunday afternoon. Distance projected always matters. Clearly for wireless the energy matters. For undersea cables, the energy matters (ever tried powering a repeater in the middle of the ocean?). For physical transport of bits, energy matters. Maybe for mains-powered land-based things it doesn't.
All a bit of fun! Enjoy finding the arithmetic and factual errors...
According to Slashdot, Google may be running up to 100,000 servers. Google's economic innovation was to capture the opinions of the general linking public as proxies for endorsement, rather than use an army of paid taxonomists and evaluators like Yahoo did. (The other search engines that solely relied on the page describing itself are already distant memories we can joke to our grandkids about.)
So if you want to do in Google, here's the next step: replace the expensive 100,000 central servers with a distributed search engine. We've already seen a scratch-my-back-and-I'll-scractch-yours model succeed with BitTorrent. Your download speed depends on your upload generosity.
An unintended consequence of the remote attestation feature of trusted computing is a strengthening of peer-to-peer networking. This feature lets me know the program running on your computer is the one I think it is, not just one that can spoof the protocol. So if you want to receive a cut of the advertising revenue, you had better be doing your fair share of crawling the Web, assembling a distributed index, and answering queries from nearby nodes. No chance of gaming the system.
Many fear the opposite: increased central control over the edge. The usual story of the technology not being evil, just the use you put it to.
Of course, Bill Gates might be even more evil than we ever thought, and in future Windows will be free, but the encryption key to use the remote attestation feature will cost you big time. But whatever happens, the current balance between the center and the edge is likely to come under stress.
While researching my previous article I came across this gem on the Air Traffic Control (ATC) industry. I thought I'd share it with you.
Stuck-in-the-muds argue that ATC is a natural monopoly, that there has to be just one provider. Nonsense. The phone companies interconnect invisibly, power companies share electricity as needed, credit bureaus and other services swap data constantly. Air Traffic Control isn't any different. Today we have a mix of civil ATC and military ATC and non-federal control towers. They work together seamlessly. There is no reason we cannot have much more.
Aviation is at a decision point. Do we go ahead into the next century with the advancements that modern technology offers? Or do we refuse to budge? AOPA, EAA, and the other alphabet organizations argue that we must not change. They want ATC to remain part of FAA. They believe we should continue to do things as they've always been done.
Sounds awfully familiar, doesn't it?
I've been meaning for a while to write about the possible consequences of Microsoft's entry into voice telephony. They're clearly serious about it. Their standard strategy of "bundle with Windows and suck the life force out of adjacent markets" fits nicely with eliminating telcos from routine business voice traffic.
The reference design for a next-generation PC was unveiled last year at the Windows Hardware Engineering Conference (WinHEC). Codenamed the Athens project (Word document), it integrates three key related features:
Clearly this would enable a PC to turn itself on and answer a phone call if necessary. Always-on comes to the edge at last!
The obvious question is whether they can carry it off with Longhorn. The PC is a very costly device compared to a deskphone. Laptops, not fixed desktops, increasingly dominate the corporate landscape. We'd all need to move to Bluetooth headsets if we don't want to keep on plugging microphones in all the time. Cell phones already lie in every handbag and swing from every beltclip. That means alternatives are already within arms reach. This market isn't the traditional Microsoft hyena picking off a vulnerable young business wildebeest. (By the way, it's worth clicking on that last link for a dose of irony.)
But let's assume that this all comes to pass. Softphones are integrated into Windows. Consumers are gently shepherded towards the MSN phone service. Businesses deploy a Microsoft SIP/PSTN gateway. The Nortel usability abomination I have on my desk goes to the landfill. What happens next?
Well, this exposes the PSTN to a new risk that it may not survive. We've already seen scumware that installs itself on your PC and dials premium-rate numbers on your behalf. Fully automated remote fraud. (No jokes please about this being the standard way telcos issue phone bills.) But doing this via a hardware modem attached to a POTS line doesn't scale well. Any user sat there will hear the clicks and chirps and gets suspicious.
So we move every Windows desktop to a standard softphone interface attached to a live PSTN bridge. Now you've done something to dramatically increase the scale of the risk. To the greatest extent possible the telcos have kept the PSTN closed. None of those SS7 signals are allowed to leak out to end users. All you get is a simple analog line with a very limited repertoire of messages and a trivial call-and-answer protocol. And of course some of the most famous moments in telco regulation have come from attempts by the network operators to control what can be attached to their networks. But you can still get a computer to initiate an outbound call. The telco control doesn't extend to how calls are initiated.
If we're not careful we're eventually going to get a Windows desktop worm that does more than just forward your home-made porn to your colleagues and snitch on your Internet banking password. It's going to make phone calls. Lots and lots of them. Silently. And along with hundreds of thousands of other zombie PCs, it will take the PSTN down with it. Five nines will revert to being an ice dancing championship score, not a network availability promise.
On top of this, remember those phones are attached to loud bells that wake people at night. We're not talking packets being filtered quietly at your hosting service when being hit with a SYN flood here. Members of the public will start to wonder if this old-fashioned phone thing is such a great idea after all.
To prevent this you would have to ensure every outbound call was mediated by the operating system and authorized by a human. Automated outbound calling to arbitrary numbers would have to be verboten. Not likely to happen, I think.
The ultimate irony is that the Internet is criticized for its spam, fraud, denial of service and anonymous miscreance. But that just reflects humanity's self-destructive streak. It can cope with a fair dose of abuse and remain operational. Once the Internet hosts are able to address the PSTN hosts, the PSTN doesn't have the immune system to cope with the assault. It's like measles arriving in the New World. It just wipes out civilizations because they can't adapt fast enough. And you'll be able to thank Microsoft for bearing the virulent strain that kills carrier-controlled telephony.