Net neutrality - the great debacle
Dominic Sweetman disputes Richard Bennett's account of the birth of TCP/IP, particularly the "end to end" principle, which Bennett characterises as a cargo cult:
TCP/IP's 'end-to-end' nature (what used to be called a @connectionless network layer") - far from being the consensual fashion of the time - was ferociously denounced by communications experts. They basically said it was just the sort of rubbish you'd expect with a communication system invented by computer people: it obviously wouldn't scale.
TCP/IP worked, probably because the computer people didn't trust the network, so they wanted something robust against all sorts of network malfunction - the DoD wanted something robust against enemy action for Arpanet, which kind of kicked it along.
For what it's worth, the points Bennett makes now about the weakness of the IP-based internet are exactly those made by communication gurus at Logica when I was a youngster there in 1978. They were wrong. Things do come around again: just because something was wrong in 1978 doesn't mean it's necessarily wrong in 2006. But it behoves anyone trying to buck the wisdom of history to acknowledge the history and get it right first.
That might be a little unfair - it perhaps misreads Bennett's point that the internet had to adapt to survive, and needs to do so again.
Bruce Johnson is one of many who argues that more bandwidth will solve the problem.
You can simply build more capacity. If the telcos can't do that selling their bandwidth for the price they're charging, well, charge more for the bandwidth.
Mike Bostrom finds it more complicated. Although he says more bandwidth fixes the problem to some extent, he agrees that SLAs (Service Level Agreements) are desirable:
What I *am* against, is that the network operators themselves are setting up an infrastructure (both physical and legal) to be the sole - or at least leading - providers of such services. That is, they would control the largest media distribution services in addition to the connection and delivery technology.
Under such a setup, they have major financial disincentives to provide equal access to the network, even if they had capacity in the form of dark fiber. We both know that most operators have such untapped resources at their disposal but inactive (How else could Google have been able to buy that same available but unused capacity?)
I have no doubts that as soon as the operators have set themselves up as the only working providers of the new services, the profiteering will start.
As a nice encapsulation of the paranoia we identified, here's David Sloane, who writes:
The problem is not engineers, or political activists. The problem is greedy executives at entrenched telecom oligopolies who want to charge more money for the same quality of service they're providing today. Bennett conveniently leaves out the fact that the tiered-Internet we're talking about involves increased cost for the same service being offered now.
That's the fear, you see. But he offers a concrete example:
My company uses two AT&T data centers and AT&T internet connections at our offices. We don't have to leave the AT&T network for these sites to talk to each other, so performance is generally good. We also talk to our clients over the internet. But we're only paying for internet access, not 'smart' internet access or Preferred or Gold or whatever.
Maybe some of our clients will upgrade to Internet Gold service with Level3, so we'll need to upgrade to Internet Prime Advantage service with AT&T. Their costs go up, our costs go up, and the service we actually use has improved not one iota. Sure, AT&T has to set up lots of fancy prioritisation and monitoring systems to support this new paradigm (and billing systems, and sales processes, and marketing materials), so they have to pay for it. But in the end, we're all just going to waste a lot of resources so Google, Microsoft and everybody else on the planet can't use the AT&T pipes 'for free'. This seems like a bad plan, for the engineers and everybody else.
Don't get me wrong - there are prioritized services which can (and should) be developed and sold by the ISPs, but these should involve some additive value over the current offerings - MPLS and video/voice application prioritization are a good examples. But they have almost nothing to do with the average user of Google or Microsoft or Yahoo! web sites. These services are only useful in private or VPN networks, where there is a common carrier for all endpoints.
Sponsored: DevOps and continuous delivery