Author: Paulo Villegas
Disclaimer: Though I’ve tried not to be, statistically speaking I’m likely to be biased. Still, I hope that the reasoning holds even when accounting for bias.
The net neutrality principle is a hot controversial issue, rich in political and economical aspects, and with strong (and sometimes evolving) positions by key actors in the Internet arena. Here I will touch the topic, but without dwelling on any core reasoning. I will rather concentrate on a very specific aspect: its associated metaphor of a dumb pipe, as applied to characterize data networks (specially referred to Internet). And I will label that metaphor (at least in its standard formulation) as misguided.
This ‘Dumb pipe‘ analogy is rather pervasive nowadays, either to use it or to argue against it. In essence it is just a mapping to the communication network from distribution networks of utilities such as energy and water. Similar to those networks, in which the carrier limits itself to transport the items involved (electrons, water molecules) from source to demand, a data network considered as a dumb pipe should limit itself to carry its items (‘information bits’) from one place to another, without paying any consideration whatsoever to what it is carrying. Any value-added service will be added on top of that dumb pipe, but with a clear separation from it, in both operation and business model.
Analogies that do not work
Note, however, that this mapping does not really hold well at the basic level: the items carried by the original and the mapped dumb pipe cannot be satisfactorily compared. In water & energy transportation, the ‘bits’ are indistinguishable from one another. So the carrier does not need to worry about routing: its only responsibility is to keep the ‘traffic’ flowing, but it does not matter the least what is the source of any of the ‘bits’ filling the pipes. That, of course, is not the case in a data network, which must ensure the right end-to-end path for each bit carried. This implies that the problem at hand is fundamentally different, so the usual dumb pipe comparisons do not really work: they are apples and pineapples. Energy and water can be dumb (though they are trying not to, see below), but communication networks need be complex, otherwise they cannot deliver the service.
Having said that, note however that there are a couple of senses in which the mapping from utility networks to data carrier does bring some kind of validity. But the net effect is to make energy transportation more like data networks, not the other way around
One is the “smart grid” concept, that is trying to “undumb” the energy dumb pipe, and make the network aware of what lies at its edges. Circulating proposals would dispose of the “energy neutrality”, enabling different service agreements with customers, including dynamic negotiations depending on the current state of the grid and its foreseeable situation in the near future (a customer could accept a poorer service with a chance of occasional degradation in exchange for a discount, and in congestion situations the energy company would allocate resources based on those contract agreements).
Moreover there is nowadays a (theoretical) way in many countries through which you could ‘choose’ the source of your energy: you can select an utility company, and in theory that changes the ‘source’ of your energy bits. The selection is mostly at the ‘paperware’ level: the actual energy transmission does not change that much (no new power wiring is done to your house). To that effect, your energy company is turned into a kind of ‘virtual energy operator’, using the infrastructure of your “last energy mile” provider (the distributor) to deliver the energy.
Apart from cost considerations (some energy companies could be cheaper than others), a criterion used to decide upon such energy provider could be the production type. Energy companies give you the source decomposition by generation type, and since they have different energy production mixes, in theory you could choose an energy source that matches better your taste (e.g. by selecting a ‘greener’ company). This decomposition is nevertheless computed at an aggregate level (company mix compared against general mix), so there is no real guarantee that the energy arriving to your house does have that composition. Electromagnetic particles are not source-stamped except in weird quantum-mechanic experiments (which might yield important cryptographic applications, but are quite far away from being useful for billing purposes).
A similar ‘smartify’ attempt goes for natural gas distribution to consumers through pipes. In theory it could also be done for water supply, although water tends to have a tighter regulation, with a single company serving your area. And water distribution is still mostly state-controlled (privatization has only been done in a handful of countries, and remains controversial). It tends also to be ‘produced’ more locally, i.e. source and destination are nearer, and has lower interconnections than gas or electricity (it’s a less connected grid).
An analogy that does work
Can we then find a better analogy to a data network than the ones coming from ‘classical’ dumb pipes? Of course we can. It just needs to be a service in which the transport layer must track source and destiny of every ‘nugget’ carried. I can think of at least two interesting possibilities for such a mapping.
The first one is the post office (kind of obvious to find). It’s clearly a service in which nuggets of information must travel between a clearly defined origin and destination, and any change in destination destroys the service.
Any Internet analogy to the postal service, though, must factor in a significant difference in procedure, namely the billing method: the usual practice in postal information exchange is for the sender to pay the costs. One of the biggest breakthroughs in the history of postal services was the introduction of the pre-paid postage stamp (mostly credited to Sir Rowland Hill in 1840 with the creation of the Penny Black for the UK Postal Service). Before deployment of postage stamps, mail was generally paid by recipients (which created a long series of problems), from then on it changed to mostly senders paying for the service; this is quite unlike the Internet (‘mostly’ because postage paid by addressee does also exist, for instance by reply-paid envelopes sent with advertising mail)
So, accepting that this comparison is fairer, can we now use current post office services as an example of net neutrality? Well, yes and no. It is true that all standard letters are created equal, and in principle the post treats all of them on equal footing: universal service is a basic feature. Fares are also almost flat rate (within a range of destinations). But note also that there are specialized post services (which carry an extra cost) over the same post network. For instance, we can have additional security assurance or increased delivery speed by paying a premium. Certified mail or registered mail are examples of “value added services” that use the same network but provide assurance (and proof) of delivery (this therefore tends to imply that standard mail has no ensured delivery, or at least a certain disclaim of responsibility by the network provider).
Express mail provides additional speed over standard delivery time, again privileging certain bits within the post network at an extra cost (low delay is more expensive!). On those specialised services, regular post services face competition from private enterprises, such as DHL, UPS, FedEx, which use alternative networks.
Note also that a universal service with flat-type fares can be forced to establish limits. A famous case is the Bank of Vernal, a building created in 1916 in Vernal, Utah. The 80,000 bricks needed to construct it were posted by mail (packaged in groups and correctly stamped) from their production site, 400 miles away, thus taking advantage of the low fares in the (then) recently inaugurated Parcel Post Service (amazingly, this was much cheaper than using standard means of transportation). We could consider this an unsustainable stress on a flat-type fare plan, and the U.S. Postal Service did indeed consider it so, since shortly after honouring the delivery of all bricks, it passed a regulation that limited the total weight that any single person could send or receive per day. I.e., a data transfer cap.
Yet another one
Our second example is a little farther apart, but it can still hold the analogy: here it is also crucially important to sent each specific ‘entity’ to a single, well-defined destination. Those entities are, in this case, persons. And the network, the airline network.
What about fares in this example? Well, anyone thinking that mobile operator pricing schemes are complicated (they are) should then turn to look at airline tariff policies for really incomprehensible pricing schemes. The only real guideline is maximization of the airline carrier profit by playing with availability of seats and costs, and imposing somehow arbitrary restrictions on lower tariffs (mostly with the aim of improving the looks of the higher tariffs).
Now about net neutrality. If we follow its most strict claim, which states that the network must not discriminate any user against another in her use of network facilities, and prohibits Quality of Service differentiation based on tariffs, then airlines present a clear violation of that principle. It is commonly called Business Class. But even in this extreme case a minimum principle of non-discriminative access to “content” (destinations) is preserved, since it is rather unusual to find a flight destination covered only by business class seats. Furthermore, business class seats get bandwidth (I mean, seat-width) privileges, but not round-trip delay advantages, since they arrive at the same time.
Finally, a relatively recent addition to airline networks is that of low cost companies. These would represent the case for absolute net neutrality, given that they treat all customers equally: no business class, no reserved seats, same no-frill conditions for everyone. Their price scheme is based on adjusting costs, trying to ensure full occupancy and cutting all extras, the equivalent of a bare-bones network (the “dumb pipe” again). But they can’t provide long haul connections (apparently the low cost scheme can’t cope up with the requirements of transoceanic links), and the ‘absolute egalitarian’ principle shows also creaks, such as the appearance of priority boarding. Apparently, absolute net neutrality is too demanding to be sustainable.
Summarizing, the main conclusions to be drawn are threefold: the Internet as a dump pipe is really not that dumb, the ‘authentic’ dumb pipes (utilities) are trying hard to ‘smartify’ themselves, and networks matching a high-level analogy to Internet do not really respect the dumb pipe metaphor.
We could conclude by renaming the Internet as an “Smart Pipe”. The follow-up question would then be to what degree a Smart Pipe needs to support some or all the principles behind the net neutrality paradigm
Now that is indeed a very good question.