That Darn Internet
If we are to have a real discussion rather than posturing we need to decompose the problem.31-Aug-2009

This is also available as via NNSquad as .

I’ve seen a lot of storm and fury on the list but little effective discussion. Defending the status quo because it is what it is is not the  real discussion we need given the vast changes in technology and economics over the past century or two.

I’m continuing to try to tease apart the issues. As with any decomposition the choices come with an attitude but I’ll try to be transparent so the attitude can be challenged.

That Darn Internet.

First I’d like to know what we are talking about. What do people think “the Internet” is and why do they think it has become so important?

In particular what policies will be more beneficial to society, the economy and individuals and what policies are dysfunctional by these measures?

We do have a set of policies in place which date back to at least 1934 if not a century before. Are policies applicable to analog communications and telegraphy appropriate for today’s digital technologies?

If there is agreement that the Internet is result of a dynamic set in motion by freeing us to focus on the relationships between the end points and telecom is all about the services defined in the middle and infrastructure funded for the purposes of supporting those services then we have two unrelated concepts. What does one have to do with the other? Why do we give the 19th century approach gatekeeper control over the newer approach?

It would seem that a funding model for a services business limits the Internet dynamic. I see this as similar to having a private road which charges a very high toll for passage. What happens when the road owner’s policies threaten the economies of the towns along the road. This seems analogous to today’s situation. How should we resolve this conflict?

Business Models

As we’ve seen business models are not sacred. Sheet music gave way to recorded music. The newspaper model is running into trouble. Charging per-search for online searches no longer works – instead we use indirect funding such as advertising.

Today’s telecom is premised on funding by selling services. This model fails if we no longer have to buy services but instead create them ourselves (or buy them from third parties) using fungible bits. The capacity for carrying bits using today’s physical transport is very large and has been growing as technology improves. Today that capacity is limited by business models that take those raw materials and, in effect, converts it into a service.

Instead of copper which can carry gigabits for some distance and many megabits over longer distances – we don’t know the intrinsic limits – we have DSL offerings at given low bits rates. Does it make sense to limit ourselves to copper locked into the service model or should we have access to the raw infrastructure? I’ve compared this with the idea of locking computers into a service model rather than giving a third party software business a chance to discover more value. Same for fiber and wireless.

What business models work for encouraging us to discover and take advantage of the native capacity of the underlying physical infrastructure? (There is also innovation in using bits but I’m trying to keep this simple).

What is the cost of path-dependent approaches as opposed to taking advantage of any path available (as in running a fiber to an apartment vs using a common bit transport0.

What models fail if we have a superabundance of capacity? What happens when those dependent on the failing models have control of the capacity?

I’ve argued that we should have policies that give us the ability to communicate with zero marginal cost. What are the barriers and/or benefits to such policies?


If we do indeed go to an infrastructure model without QoS and the only option is to add capacity rather than favoring any particular service what problems does that cause? We’ve seen latency go down as capacity increased. If the bit rate is much higher than necessary for a particular video service then buffering would work very well. There would be no guarantees but then the current system offers no guarantees either. In fact “best efforts” is not entirely new – we can view statistical multiplexing as an early form.

What problems would arise if, by policy, we made such an infrastructure available as a baseline at zero marginal cost?

We wouldn’t prohibit companies from building their own but would very strong favor using the common infrastructure. When/why would a company build and maintain its own infrastructure (other than locally)?

For that matter do we need networks as such or will networks arise from our ability to do our own networking using local facilities that interconnect using common protocols without limiting ourselves to existing protocols? Note that we don’t need to use “bucket-brigade” approaches – we can have spanning connections. If there aren’t networks – just us using local facilities, what kind of funding/business models are feasible? for answers to these and other exciting questions.