My cognitive dissonance with Network Neutrality

[Original Post date 7 Feb 2016, migrated from Ubedge.com]

I am silent on Network Neutrality as I am split on the side to take irrespective of the hat I wear. Understanding the underlying science makes it harder.

Entitlement of Ownership of Network that feels like a public good

Network neutrality is a battle of ideas about the ownership of the network. It is a battle between the owner, license holders, consumers & producers of the network.

Internet is an important public resource for entire humanity, the scramble and fight for it thus is not puzzling. And this has led to frenzy amongst regulators, law makers, industry lobby and many other stakeholders.

Network neutrality ‘for’ argument is that users should get fair access to the network on fair terms and fair price. This ask for fair access makes sense, challenge though is in modeling internet akin to dedicated a physical resource to determine fairness of use.

Is there a science to ownership of network ?

Traditional network reserve a channel for information transfer for a specific time. This network type is Time Division Multiplexing.

Bandwidth, a measure of maximum throughput of information flow constrained by the low & high end frequency of the channel.

This design makes network neutrality ask possible, reserving the channel for those who paid for it.

Fig: Time Division Multiplexing Network, eg Landline Telephone Network

A simple analogy to understand this is like train & its track . The track is like the channel that is reserved for a specific time window & switches based on destination. The information is like the ‘train coaches’ that pass through the track.

Fig: Analogy for a Time Division Multiplexing Network

For internet (statistical multiplexing), there is no reserving of a channel. Many source are trying send information across the channel at same time. Information is resent if does not reach the destination.

Bandwidth, a measure of maximum throughput of information flow is a function what other information is competing to get in and the loss in the channel.

Fig: Statistical Multiplexing, eg TCP/IP network (i.e Internet)

An analogy is that of cars on a highway that have to signal the starting point after reaching destination else another one is resent.

While there are good mathematical model for computer (a Turing Machine), traditional network (Nyquist Sampling Theorem)

There exists no rigorous mathematical model for statistical multiplexing (internet). Few folks have postulated one recently have established it to be as rigourous as Turing machine for a computer.

In this model throughput of the network, loss and delay are key properties of the system and related.

It states that networks are systems that trade throughput, loss & delay and conserve the three, ie increase of two will reduce the third.

Called as theory of information translocatability or DeltaQ connects the above three parameter to the quality of service.

With this theory it is hard to bandwidth as a finite physical resource.

Net Neutrality ask here is like reserving the entire highway for my car because I pay road tax. This goes against the very reason this network design pattern used to solve channel access problem. If followed to its logical can collapse the entire design and this type of network.

A PhD dissertation explains it in detail here

Pioneers of TCP/IP network design such John Day RINA, Cook Report have pointed this out for ages.

If this theory is true, it says focusing only on one stakeholder’s ask we will create an architecture instability. We will create a network that will fall under its own weight. Is that what we want by arguing for network neutrality ?

What are things that feel right to me ?

That there should be openness for innovation in platform like internet. Much of progress in internet has happened because of that. Monopoly is a fact of life but monopolizing is criminalizing for the society. Giving current power bearers more tools to twist with more power does not feel right.

Thus the argument of keeping the distribution as neutral as possible make sense to me.

However if the theory is true we can’t change the objective reality of a statistically multiplexed network.

Thus it feels as if stuck in a no man’s land of Kuhn’s paradigm shift. If there was a Feynman it would have made things simple to atleast understand.

Summary

SaveTheInternet team in India has now framed the network neutrality as equality of competition (of opportunity & not outcomes) as that would be create maximum societal benefit. I have sided this for protecting that openness.

Disclosures: I am a contributor in many organizations, co-founder of Bangalore chapter of Mobile Monday (2006–2013, not active now). I am a Fellow at iSPIRT, software product industry think tank since it began in 2013. I also joined the general group at SaveTheInternet a few weeks ago. My day job is in leading Product Management for Quickbooks Online at Intuit India.

All above opinions are mine and have no bearing on official position taken by any of these organizations.