Mark Goldberg


Taking a neutral position

In my Twitter feed, Libby Jacobson linked to an interesting read about net neutrality: “Taking a Neutral position: How you spot an arts graduate in a tech debate“. The article by Andrew Orlowski in The Register was subtitled “Sorry lawyers, but the Packet Pixie doesn’t really exist”.

As the article indicates, “Earlier this month Ofcom quietly published a fascinating technical study which helped highlight this: casting light on the difference between the arts grad approach to network management and policy, and the hard numerical reality.” The Ofcom study, “Traffic Management Detection Methods & Tools” [download full report ], raises some important questions about the tools that many industry observers use when complaining about inappropriate traffic management.

It is inherently impossible to detect directly the specific application of differential treatment (other than by inspecting the configuration of network elements). Even when there is such an intention, it may not have any effect, depending on the particular circumstances of load, etc..

The report’s conclusions are important:

The success of packet-based statistically-multiplexed networks such as the Internet is dependent on sharing resources dynamically. This dynamic sharing is ubiquitous, occurring at every WiFi access point, mobile base station and switch/router port. Each of these multiplexing points allocates its resources in response to the instantaneous demand placed upon it, which can typically exceed the available supply.

Poor performance may have many causes, including the overall network architecture and topology, capacity planning and in-life management. ‘Traffic Management’ is only part of the equation.

Presumably for this reason, traffic management detection (TMD) has been pursued almost entirely from an academic perspective. Given the complexity of the relationship between desired outcomes and actual behaviour, inferring an intention from observed outcomes is effectively impossible. Rather than trying to address this general problem, most TMD starts from assumed intentions mediated by assumed particular TM techniques and then attempts to deduce whether or not certain observations are consistent with such assumptions. However, even positive results do not prove a deliberate intent to introduce bias; given the overall complexity of relating intentions to outcomes, demonstrating a differential outcome does not demonstrate an intent to produce that outcome.

The Ofcom report concludes, “no tool or combination of tools currently available is suitable for effective practical use.”

Orlowski says, “‘Net neutrality’ turns out to be a completely inappropriate metaphor to describe the physical reality of packet networking. Such awareness of the physical reality is what separates network techies and engineers from the typically arts or social-sciences qualified advocates who campaign for ‘net neutrality’.”

The articles are important reading for regulators and advocates on all sides of the debate.

Comments are closed.