Martin Geddes on The Future of Telecoms and Broadband - Part 1

One of my most favourite thinkers in the telecoms space is Martin Geddes who is a chief analyst at STL Partners. Martin also has a very well read blog humorously called Telepocalypse.

Because I see very strong value in Martin's opinions I asked him to keynote and to be on the advisory board for eComm 2008. And yesterday since I was in Edinburgh (Scotland) I decided to go round and do a pre-conference interview in person. We decided to aim for fifteen minutes but soon after hitting record we found that an hour and fifteen had passed. Now the word on the street is not to post interviews longer than the talk otherwise people will opt not to come to the conference! But I find it hard to believe that people could be so short-sighted, so I've decided to upload the whole thing. I've split it into two logical parts and will post the other half as a separate post.

The run time is 54 minutes.

Below is some text to give you an idea without listening to the audio interview but it is by no means anywhere near a transcript nor full account of the interview (so please listen to it). Rather it is taken from notes I made in relation to points that interested me personally. I've also added hyperlinks where I believe they could be beneficial to understanding.

I started by asking him to tell me about the Telco 2.0 initiative to which he responded:

Telco 2.0 is about moving to a world where vertical integration of the telecoms industry is coming to an end and your seeing new industry structures emerge and it is how to remain profitable and alive in that environment

He made it clear that it is not only the applications themselves (voice and SMS) which are being horizontalized but the industry itself in a macro sense.

Martin also spoke at length about varying degrees of both vertical technical and commercial integration which results in different patterns of systems for distributing data to users. One of the examples he gave was:

...downloading a movie to watch on a iPod has a low level of technical vertical integration - the ISP does not know about the movie. But Apple may be paying CDNs to cache it, so there is some degree of commercial vertical integration. When looking to deliver voice, video and data, you can adjust the two axis of commercial and technical vertical integration

He went on to say that separating the application from the delivery mechanism which one would normally assume is a good thing, is not always good:

There is an utter total fixation on the broadband ISP product which is one where the applications have no access to the means of transferring money based on who is going to pay for delivering this movie, there is no API for hey Mr ISP...I will pay for it's delivery, don't count these bytes towards the users monthly usage cap, which is what the application provider would like....there are benefits from separating the applications from their delivery but also looses a lot in that process of the commercial richness of telephony and SMS and that is a problem

He then made a pertinent statement about the future of the Internet:

The future of the Internet and the future of broadband is largely about putting back into it some of the rich commercial models from other distribution and delivery systems like the fixed telephony PSTN system but without all the technical limitations of those old legacy systems

When I homed in on the net neutrality issue with the question:

you decide how many gigs you want, what speed you want, you purchase connectivity as a product, surely that should not be bound to the application, the applications you may or you may not pay for but these are separate things?

Martin responded:

there are a number of subtleties to that question...when we say connectivity what do we really mean? And there is two parts to that. One of which is getting the pure access part...how do I get from my mobile device to the cell tower...separately from that there is the question of getting the data across the world to wherever it needs to go. What we are seeing is the access component is relatively steady over time...the users do not like having to think about how many megabytes and megabits they need. So if you want to access the new BBC iPlayer online and your on Tesco value broadband, you've only got a 1 gigbyte usage limit in the month, the average user is not wanting to have to start thinking about the bit rate of the streaming video. All they want to know is either I am buying some product and it comes with postage and packing included or it is an add funded thing from something like YouTube and Google are paying for the whole delivery...and also they don't want to think about having to provision a new access everywhere they go, so when you go to your parents for Christmas...you really don't want all your usage at your parents house to be counted against your parents ISP plan because they might be on Tesco's value ISP broadband plan and not on the super business class version that you prefer. So we have to go work out how to package up the access, the data service itself and the application of devices in new  and different ways to sell the user exactly the thing they value and no more and no less. The best example around at the moment is Amazon's Kindle...this device will deliver you ebooks, you don't have to think about megabytes and megabits or whether it is provisioned, or hotspots...it just does it...it has been packaged up perfectly.

In relation to another application which is bound to the device and connectivity Martin states the SMS market last year was $100 billion dollars which is bigger than movies, music and computer gaming all put together.

for the few gigabytes worth of traffic that all of SMS adds up to, some small amount the users are paying some very very large amount of money but only if you view that purely from a very technical viewpoint. What that is telling us is a couple of things. One of which is that  when you brake apart a product into small pieces and sell it by the sachet rather than by the big bulk bottle,  you get much higher margins....another thing is that SMS is a distribution system for messages and data. And each of these distribution systems has a large number of attributes. One or two of which are what kind of technical data can it carry and how can you make payments thru it but it has a whole load of other attributes as well so for example the difference between SMS and the Internet is that SMS is based on phone numbers, phone numbers are attached to countries so for example if you are running the Eurovision song contest and you want people to be able to vote for songs, but they can not vote for songs from their own country well with SMS it is easy to filter out those votes....with the Internet you can not know with high certainly which country or jurisdiction  and ISP address is attached to and it is easy to go and fake it as well...it is only by understanding the full suite of attributes of these different distribution systems  and their cultural context and their expectations like telex has particularly laws, it has legal force in a way that other forms of communication maybe don't; fax only became a legal form of communication 10 or 15 years ago even though it existed for longer. So you need to look at the whole context of each of these different systems. The clever thing you can do with a telco or an application provider is blend together these different systems to create in effect new communications media. The impressive example at the moment is something like the Sky Anytime Box...it manages together broadcast and broadband...there are multiple ways of delivering the bits to the user...Sky have blended together the best of different distribution systems into a new one.

Ingeniously Martin has been thinking of the Internet as a means of signalling and coordination rather than always also the best means of delivery. Martin also steps into heretic waters by knocking the fixation with VoIP as a means for moving voice:

Netflix is using the Internet as a signalling system and the postal service as a bearer. And the postal service is a very efficient way of transferring tens or hundreds of gigabytes worth of data....the important lesson is that when you take this to where the cash is - the money is in voice - is that there has been this fixation with voice over IP for a number of years and actually, maybe and this is heresy but maybe the good old telephone system is really good at transferring voice. Hey time division multiplexing, does constant bit rate voice, real well! So you have to throw an awful lot of [VoIP] technology at a problem [voice quality/delivery] that does not exist to try and persuade anyone to move over to voice over IP. So it is only by understanding the full context and capabilities of each of these systems that you start to think hey actually the Internet is good at allowing new forms of signalling to evolve faster than what SS7 or whatever may have allowed...so why don't we focus on allowing the IP part to do what it does well which is how do we enable the rendezvous' in front of this phone call, how do we return signals and presence data and the little picture of where I am at, location information to help people make phone calls at the right time...stop worrying about trying to do voice over IP until the technology is super duper mature until we can not possibly afford to maintain two networks which is quite a long way away still and let the phone network do what it does well which is phone calls.

 When Martin said that I could not help but ask what he felt about 21CN: 

...if we just focus in on the voice part of the network itself...what I'd be tempted to do is look at how long from inception to delivery [of 21CN]...that is an awfully long time to be planning a technology project over, in this environment, and you could have built a wrapper around the existing voice network...and started to enabled lots of the "advanced services" that the 21CN would offer, years earlier, if you wanted to...you can see the allure of moving to an all IP network but you also have to understand the high risk of having any project that has 5-10 implementation time scales, where the protocol and architecture you are choosing may be obsolete by the time that you get to the end

SIP has been facing growing criticism and Martin picks up on that to prove his point (as SIP is a core protocol in BT's 21CN):

 if you look at something like SIP which was the obvious clear winner out of the IP signalling protocols, even that has got it's problems, it's got it's architectural mistakes, it's incompatibilities. It has been an absolute nightmare to get everyone to have an understanding of what a particular SIP message means...as a result the chances of  SIP itself being exposed to the application developer and being the interoperability layer, probably is not going to happen. SIP and IMS is good at provisioning the signalling to say hey Tom over here has paid a certain amount of money to have certain kinds of traffic being hauled over this network...it is good at saying hey lets manage some virtual circuit. What it isn't necessarily good at is trying to have a common understanding of what some presence message might mean in the future because it's all contextual to the user

Martin redefines what telcos are which is neither a dumb pipe nor an application provider:

let the signalling protocols do what they are good at which is help the telco be the logistics provider for data...personalized logistics providers for data, that is what telcos are. They haul data around from point A to point B but they do it using a number of means, so could be broadband, could be content delivery networks, edge caching. The telco provider is just like a logistics provider for atoms, blends together rail, sea, land and air transport...for voice you have prioritized networks vs best effort vs full circuit switched, packages it up with the applications and the device but does not get involved in the application itself. All they are doing is helping these people get data A to B, which isn't just being a dump pipe, it is a lot more complicated that just being a dump pipe because your having to slice and dice up that pipe and enable competing, contending uses of the pipe to be prioritized against each other to have much finer grained provisioning of the pipe...

When asked in relation to the electricity analogy - "is this not like saying that if you buy a kettle, someone else should be paying for the electricity that goes into the kettle" he responded:

When you post something there are all kinds of ways of deciding on who pays for the thing being posted, could be the person at either end....electricity distribution is like broadband and doesn't offer these different payment mechanisms and there are good reason for this. It could be good that each appliance has to account for its own usage. There are good reasons why the power network does not have this capability....power distribution is vertically integrated for a very good reason.

 Since it was clearly apparent that at least on the surface those positions seemed to differ extensively from what network neutrality folks appear to be advocating I steered the conversation closer and closer to the heart of that debate, to which Martin responded:

the whole network neutrality debate should your ISP ...price discriminate and reclaim all the value for themselves...there is a duopoly in the US and there are some pros and cons of that model. There are some lucky people in that model where Verizon FIOS has been deployed because the Verizon model says this is a zip code with higher income people who pay their bills...it works to some degree....it [network neutrality] will also greatly inhibit the growth of the necessary economic models that will be needed in the future, particularly the growth of rich wholesale markets in access and connectivity.

 Martin goes on to say what would be termed anti-network neutrality is in fact a good thing:

The vertical industry structure you see in the USA, where AT&T and Verizon and a few other companies dominate the market and you have limited retail choice, where you have the choice of one or two broadband ISPs. In the UK market there is a thriving, vibrant wholesale market. In France a thriving wholesale market, a foundation in which to grow new wholesale products which would enable new kinds of connected appliance, so you go out and buy the Apple TV box and plug it in and you don't have to think what broadband plan your on...the iTunes service has to worry about getting the content onto the box and paid for it. So as long as you do not have a bottleneck in that little access loop stopping retail competition then the network neutrality debate, this whole bogey man, is a shadow on the floor. The shadow is something you want, which is the rich wholesale market. The idea that Google will be charged extra for delivering YouTube videos, it's, here is the secret, a good thing! It is a good thing that Google could offer you YouTube videos and you can watch them all day, in standard def., in high def., and Google is footing the whole bill for you, which today isn't the case.

Due to fear of loosing that which I value (access to Internet sites without my telco/ISP trying to rob me on the way) I asked "excuse my instant pessimism, you will get the same Google video quality but the telco will charge you 5p or 10c everytime you watch a video?" Martin replied

the worst fear of network neutrality is that your ISP will be charging you to go to something that you value highly. That is'nt going to happen. It is'nt going to happen.

When asked why he was so sure the response was:

the charges are going to be to those upstream parties like Google, to people like Microsoft... it [network neutrality] is trying to treat the symptoms of a lack of competition caused by the access bottleneck. You can not cure cancer by giving people more morphine. It doesn't work that way. It seems to take the pain away but they still die. In the US market they need to understand that the infrastructure is not part of the telecoms industry, basically. The infrastructure is part of a more multi-utility access business and their job is synchronize up the replacement of the gas pipe with the laying of a new fibre...

Martin laid a further attack on network neutrality advocates stating:

if you want network neutrality, neutrality has to be a difference layer. It is not the IP layer, it is not layer 3. Neutrality has to happen at layers zero and one....They picked the wrong place to go and fight the battle. The good news is that if you started to initiating anti-network neutrality legislation, I don't think anyone could actually write anything down which would be meaningful and enforceable. So it is largely hot air and time not well spent.

 Knowing that STL have been working on a future of broadband report, I asked to know a little about it:

we've just finished a 6 month study on the future of broadband, future business models of broadband and to do that, we did an online survey which I know 800 respond, we've talked to dozens and dozens of people...we've been comparing broadband to the other industries like container shipping  and the power distribution industry  and we've been trying to learn the lessons we can from these other industries, what does it mean for broadband and our conclusions are very much along the lines of what I have been describing. Broadband is part of a multi-network data logistics business, it does not exist in isolation

Martin made the point that in the future we should expect to see a closer relationships between the applications and the networks in question:

the biggest credit checking company in the UK, is BT, yet if your running an application on a BT DSL line you have no means to interact with that personal private data. Location and presence are other examples...the telco or the broadband service provider is a provider to these people building these applications on top of the network and doing things which are natural by products of running a network, the logistics business

 When asked directly "is your position on network neutrality, that it is a moot point", the response was:

pretty much yes, because I've never managed to see a really meaningful definition of net neutrality. I've saw lots of high level superficial definitions but when it comes to applying those ideas in practice, there far too wholly and vague. So it's just a name for a fear rather than a viable policy that anyone in regulation can adopt. You can think about has any network been neutral?

 Back on the topic of ISPs he stated:

the current ISP model tries to charge everyone the same amount for a very broad spread of actual usage, you can not remedy that though by going back to metered usage as people hate having the clock put on, they don't understand the megabytes and megabits stuff and you cannot packet shape your way around the problem because the users don't really understand advanced network policy enforcement...the way to solve it is to take the voice and video traffic, take it out of the ISP product and do it in a way that is beneficial to the user...

 Making his point that network neutrality is just a name for a fear and little else:

the network neutrality fear is that his bogey man is going to come along and rob you. The bogey man really just is a shadow of something we actually want which is a rich wholesale market.

 When challenged with "but you said right at the start that there is not such a rich market in the United States, so the network neutrality fear is therefore very real then, surely?" Martin passed comment on the future of the US broadband industry:

It's real in that there is a gap between what people are being sold at retail by an ISP and what they are actually getting. Comcast can suddenly come along and decide they do not want to have this peer-to-peer traffic even though they sold you Internet access but the solution to that is to restructure the industry, so users have retail competition and have lots and lots of people trying to compete for their business and whoever turns out to be a two faced liar, you just don't renew with them and go with somebody else. Now that is not going to happen within the current US industry structure and the consequences are that probably the US will fall behind as an industry leader 

Finally I could not help but bring into the interview a topic which has been very much on my mind by asking "have you given any consideration to open wireless? As in the 700Mhz type stuff?" to which Martin replied:

there are certain benefits to vertical integration in that it helps to pull more money into the infrastructure layer, you cross subsidize from your over priced application services because you can control what runs over the network. The problem then is that you end up with a very stagnant sort of services network because everything is gate-keepered by a telco and the telcos idea of progress is not the same as everyone else's. Superficially open access is very attractive but it is really unclear how the contention between the commons of the network and the need for innovation and the need to ration out a fixed amount of space is going to work out. So I am really keen to see the experiment and I think the experiment should be well run. There should be enough spectrum given out over to open access rules. There is a number of different models, potentially, so something like WiFi which is a bit of free for all so long as your power level is not too high. Some where it is open access, where it is all wholesale and business to business type transactions. So lets see the experiment run. I don't think anyone is sufficiently clever to understand all the variables involved, so let's see the experiment run in many different places as I don't think anyone knows what the outcome is going to be.