L'Ombre de l'Olivier

The Shadow of the Olive Tree

being the maunderings of an Englishman on the Côte d'Azur

20 June 2006 Blog Home : June 2006 : Permalink

Net Neutrality

I've seen lots of these http://www.internetofthefuture.org/ ads floating around so I thought I'd click on one to see what it said. I don't think they really wanted me to do that though... because it did change my mind - in favour of net neutrality.

As a free marketeer I was in favour of letting the telcos try and foist their version of a network on us because I was pretty sure that they'd lose. However having seen this flash presentation I have decided that they are in fact so wrong that actually congress ought to intervene and tell them to STFU, although I still think that congress probably ought not to regulate them other than that.

The idea of "internet of the future" is differentiated services, that is to say some wise people in the telcos decide which sorts of traffic are more important based on some magic identifiers (trust me you don't want me to explain this bit), and then they mark packets for these sorts of traffic when they start so that they go on optimized routes through the internet. It is, if you like, giving them a pass to the car pool lane of the information superhighway.

The information superhighway metaphor is a good way to understand why the telco idea is brain dead. In the physical world when a highway begins to suffer from congestion there are choices. You can either try to ration access to it (tolls, car pool lanes...) or you can widen it by building more lanes (or as a related alternative you can build a separate highway near it that covers much the same ground). Keep that in mind while we look at the telco claim. They are claiming firstly that there is congestion (or will be if they introduce all these sexy new services). They have to because if there were no congestion then no one would want to either ration access or widen the road. You may have noticed that many car pool lanes only operate at "peak periods" and there is similar reasoning on the internet. If you have 50Mbps of traffic in a 100Mbps pipe there is no congestion, its only when you start getting close to 100Mbps of traffic in the 100Mbps pipe that it starts seeing congestion.

So what happens then? well you could upgrade the capacity of the routers and switches on the affected link(s) from 100Mbps to 200Mbps (or even 1Gbps) or you could upgrade the routers and switches so that they prioritize packets (as in turn on the car pool lane).

In the physical world of interstates highways, car pool lanes (and the like) are frequently necessary because there is no way you can widen the road and no place to build another highway, not to mention the fact that building one takes a while (has Boston finished the Big Dig yet?). In the online world there is no such excuse, a single fiber can take terabytes of data (I think the world record is something like 2-3 Tbps but it could be higher) and, while terabyte capacity is expensive, upgrading from 155mbps to 622Mbps to 2.5Gbps to 10Gbps is not so unaffordable and today many links run at the slower 155Mbps and 622Mbps rates. In other words the valid excuses that transportation planners have that force them to go for rationing access through metering and car pooling simply don't apply online.

Oh and did you notice that I said that enabling prioritization requires upgrades? It may not be as expensive as installing higher capacity line cards but actually configuring a network for prioritization is not cheap either and could (depending on circumstances) end up more expensive than simply adding bandwidth. It is worth noting that many enterprise LANs have gone for the higher bandwidth approach because the cost of bandwidth has dropped so much that where 1Gbps was expensive 8 years ago and 100Mbps ridiculously expensive 16 years ago; today 1Gbps costs about what 10Mbps did 16 years ago. In other words rather than try to figure out which applications need the bandwidth it has worked out far cheaper in LAN environments to simply increase the size of the pipes by an order of magnitude or two and make the problem go away.

So, given that adding prioritization is not a cost-free exercise why are the telcos so keen to prioritize rather than throw bandwidth at the problem? Its obvious isn't it? they figure they can make more money by charging a premium for access to an artificially starved network. Its like a post office that deliberately throws away 10% of second class mail so that it can charge a premium for "guaranteed delivery" first class mail. If you wonder why people are salivating at WiMAX and 1001 other wireless alphabet soups this would be why. The telcos have got themselves a near monopoly on bandwidth and they would like to reap what monoploly profits they can from their monopoly.

Originally I was content to let them try and exploit what they think is a monopoly because I don't think they realize that their monopoly is on rather shakey ground. At present it isn't quite worth it for electricity companies, water utilities and other competitors to offer retail services even though they mostly have the potential to do so, but if telcos push the price of bandwidth up enough then it does indeed become interesting (currently most utilites offer fiber in their conduits wholesale to telcos). It is also entirely possible that the portals and ecomemrce sites could form a cooperative to buy alternative access. Yahoo Japan runs its own ISP, and it cooperates with BT in the UK. In the US recall that Google has bought up rights to a lot of dark (unused) fiber. I'm not clear what the US rules are for "local loop unbundling" but it probably doesn't matter. It would not be diffucult for some combination of eBay/Amazon/Google/Yahoo/Microsoft/Apple... to set up an ISP that served consumers and businesses in the 100 or so major US metro areas and offer 10 or even 100Mbps bandwidth to subscribers for $10/month. If they did so what do you think the value of telco stock would be?

I'm still mostly of that laisser-faire opinion (as is Andy Kessler at the Weekly Standard) but I think I want to invoke a minor public interest defence and say that I would vastly prefer it if the bandwidth providers and content providers were separate and hence I would prefer it if the telcos remain in business. Thus for their own good they should have net neutrality imposed on them because if they get their way they will be out of business unless they show more competitive awareness than they have in the recent past.

Update/addition: I see from the Radioblogger transcript of Hugh Hewitt's interview with Mike McCurry, one of the supporters of the Telco group is the vendor Lucatel (a.k.a. Lucent/altatel), a vendor who has a strong relationship with the incumbent telcos. Mr McCurry uses this to claim that the vendors support the lack of regulation approach. Now it is possible that other vendors support this approach but are not on the list but I can't help but notice the conspicuous absence from the list of companies like Cisco and Juniper who provide much of the newer internet equipment and who might be expected to be in favour of the telcos if it really were as clearly sensible as Mr McCurry claims. My guess is that Cisco doesn't mind whether it sells bigger pipes or packet filtering, but Lucatel wants to sell the latter because it will make more money doing so.

I despise l'Escroc and Vile Pin