The Federal Communications Commission says it will take up the matter of new "network neutrality" rules at its Oct. 22, 2009 meeting.

 

One would hope more network engineers and architects will be centrally involved in the deliberations, instead of the usual cast of attorneys. There is a very simple reason: duplex communications networks are fundamentally different from broadcast networks, and network management might not be getting as much attention as “policy” management (meaning the politics of creating new rules).

 

Duplex communications that support any-to-any communications that go both ways, symmetrically and simultaneously, are quite different from point-to-multipoint radio, TV, satellite, cable or other networks designed to "multicast" (deliver one copy of something to millions of users at the same time).

 

Any-to-any communication networks of any size are expensive, especially when they are expected to be used by "most people," as voice networks--both fixed and mobile--and the Internet access networks assume.

 

Networks could be engineered for absolute peak traffic so that no calls or sessions ever would be blocked. But those networks would be so expensive they could not sell service to most users at low costs.

 

So engineers use statistics to figure out what the peak minute of the peak hour of the peak day of the week is, and build networks that normally allow for reasonable performance even during times of peak load.

 

"Reasonable" normally means actual blocking of outbound call attempts of between two to five percent, at the peak minute of the peak hour of the peak day of the week. During extraordinary conditions even that level of blocking can increase dramatically. 

 

The point is that, in order to keep costs within reasonable tolerances, engineers also assume some blocking of voice calls, as a standard engineering practice. The alternative is to build networks so expensive that retail costs would have to very high, even if investment capital could be raised.

 

The same sort of thing is done by engineers who build Internet access networks. Not everybody with access is going to be using the network all the time, so the trick is to build networks that work reasonably well at times of peak load, but still allow for sharing of expensive network assets.

 

That is necessary so the amount of capital is limited and retail prices for end users can be kept far lower than if networks had to be built on a completely non-blocking basis.

 

A rough rule of thumb for Internet access networks is sharing of about 20:1. In other words, enough capacity is provisioned for actual use of one customer out of 20, at any given moment. What would happen if all customers actually tried to use their connections at the same moment? Access would slow to a crawl.

 

So what do ISPs do when loads start to increase? They reduce the degree of sharing, lowering the resource sharing ratios to 15:1, for example, or perhaps 10:1 for higher-value business customers.

 

"Blocking," in other words, has been sound engineering practice for more than 100 years, for some very good reasons. Nobody can build a network that offers low end user costs, is capital efficient and never blocks access, at least when that network is expected to be used by most people.

 

Some might argue that non-blocking access is possible if network use is quite light (not many people use the network, and those users put light loads on network resources). The problem is that any broadband network with light usage is going to tend to have light revenue as well.

 

It is a reasonable public policy goal to prevent discriminatory behavior by actors with significant market power. But it is quite another thing to confuse that goal with the equally-important goal of building high-capacity, continually upgraded broadband networks that offer low retail prices.

 

It appears that for both goals to be accomplished, we will have to have lots more engineers and architects involved in the policy formation process. The reason is simple: efficient networks offering low cost block access.

 

If new net neutrality rules are about full "non-blocking access," networks will suddenly become much more expensive. Retail prices also will become much more expensive.

 

Granted, there is a difference between anti-competitive blocking, for business advantage, and blocking policies intended to ensure reasonable network cost and retail pricing. But the two might be directly related if policymakers are not careful.

 

There are too few engineers now involved centrally in the process.