The Norvegian operators have agreed on how to deal with the issue of Network Neutrality and enumerated these three principles
- Internet users are entitled to an Internet connection with a predefined capacity and quality.
- Internet users are entitled to an Internet connection that enables them to
- send and receive content of their choice
- use services and run applications of their choice
- connect hardware and use software of their choice that do not harm the network.
- Internet
users are entitled to an Internet connection that is free of
discrimination with regard to type of application, service or content
or based on sender or receiver address.
I think they are good, but I think they fail to fully capture some other important aspects, which I try to explain below.
Why there's an issue with Net Neutrality ?
We are discussing about Net Neutrality because all traffic is
collapsing on one IP network and latency sensitive traffic needs to be
protected by excessive resource usage; while latency sensitive traffic
and IP best effort traffic traveled on separated networks, there was
little issue (except clear anticompetitive behaviours) for traffic
management. In those times the discrimination was total: "premium
traffic" on one wire, best effort on the other.
I
have seen many variations in the possible definition of Network
Neutrality; there are many angles the issue can be looked at and, when
you get into details, "net neutrality" tends to mean different things
to different people.
Three classes of motivations for traffic management
IMHO, the three main possible reasons I see for traffic inspection and/or management over a network are
A) to review content flowing on the network in search of specific kind of traffic and take consequent actions
B) to manage traffic in order to optimize throughput of specific applications
C) to manage traffic in order to favour applications provided by specific entities
Is content filtering useful ?
A) can include bans on
specific contents issued by some authority, something that, if not
supported by adequate guarantees in the legal system, could lead to
censorhsip.
I am not going to argue about civil and political rights, which, IMHO,
are the most important reason why any citize should oppose content
inspection; I'll try to focus on technology alone.
Inspections have always been carried to be at bottlenecks between state
borders. If there's nothing to inspect because nothing is inspectable,
the border does not exist anymore.
It is common to hear politicians asking if this technology can inspect
traffic and technology providers argue that it does. First of all it's
present kind of traffic and secondly it's the wrong question.
They should not ask if it works (indeed you can evaporate wter, it
works!) but if it can be a mean for enforcement (you cannot solve
tsunamis by evaporating water). The right question is "how difficult is
to circumvent these filtering ?" and the answer is obviously, not at
all.
We must understand that the growing power of electronic devices puts
strong criptography into the hands of consumers and these technologies
are unstoppable, thanks to electronics and furthermore are needed in
order to ensure authentication of users, corporate VPN traffic, digital
signatures, etc.
The side effect of these idea of content filtering is therefore that, if such practices were widespread, it
would become an incentive for applications developers to build
applications that put the power of strong criptography in the hands of
the average Joe to "defend" themeselves, something that would cause the
hiding of the traffic. An effect clearly against the interests of those who propose content filtering.
Network optimization and the users' point of view
B)
can include prioritization of time-sensitive applications (such as
internet telephony, videoconferencing, application sharing) or
degradation of traffic which might be considered "harmful" for the
network carrying it, such as peer to peer traffic. The level of
management needed depends on the available capacity of the network as
users compete for a limited, rival resource. If network capacity was
larger than the sum of the capacity of the local loops connecting the
users (e.g, if the capacity was not a rival resource), the management
of traffic would be limited to the end user's local loop in order to
ensure the grateful flowing of time-sensitive applications' data. As
backhauling and core networks will increasingly rely on optical fibres
and optoelecronics prices decline, the backbone will tend to be a
nonrival resource and this kind of traffic management need will
decline. Actually, the backbone capacity needs to be far less than the
sum of the capaicites of the local loops, as there is a limit to the
amount of traffic an endpoint can consume; this seems to be the case
considering the slowdown of traffic growth in certain economies where
fibre networks are widespread.
(B1) Operators tend to think
of network management as a way to save them costs, rather than as a
service to the users. In this way, they look at traffic management as
something they decide and enact that customers have no option but
getting it. Another possible way would be to offer traffic management,
on user's request, with a differentiated pricing. Personally, I am
paying a premium to use an ISP which prioritizes VoIP traffic so that
my voice calls don't get crippy. Following this schema, I think there
is a latent need by many users to demote their low priority file
transfer traffic, promote the web traffic and have a premium VoIP
traffic. Why not trying selling quality of service as an additional
feature users might request rather than imposing it to all users ?
Traffic management and its anticompetitive risks
C)
this behaviour can be enacted to benefit the same operator's
applications or some operator's "friend"'s applications. In the first
case there's a point for an antitrust case (as the operator owns an
essential facility for those users connected to his network), and even
the second cannot be as evident. This kind of management does not
require to be specifically technical; it could be based on pricing with
a price structure designed to discourage specific kind of traffic. For
example, a pricing that would include a high session-setup fee as we
find in wireless data plans, discourages the use of sporadic, short
data sessions, such as the traffic required by messaging, thus
protecting the operator's SMS business. (*)
(C1) I have been
personally (successfully) envolved in a case whereby an incumbent
operator argued that a new IPTV service he provided to his retail
customers was a "new service" encompassing a vertical integration
between data transport and the TV content. The operator argued that it
should have not been looked at this service from an antitrust
perspective related to the data transport, but rather looking a the the
TV market, where the operator had a marginal presence. The point I
brought to the court was that, if a competitor wanted to build a
similar service (with on demand content), he would have needed to buy
from the incumbent the data transport with similar characteristics and
thus even that layer should have been exeamined and be replicable to
other possible competitors. This case clearly shows that when "new
services" come into the regulatory arena whereby an operator argues
there should be a vertical integration, it is not that obvious that a
court would rule in favour of an "horizontal unbundling" of the service.
(C2)
One other case I have been envolved with was an incumbent operator who
intended to benefit from network effects providing his customers a
higher quality service (videotelephony) refusing to interoperate with
other operators at the peering points (and even not releasing the codec
and signalling protocols) in order to attract more customers. When
formally requested for interoperability with the same quality of
service at the peering points and full details of the signalling, the
operator preferred to withdraw the service from the market. (this leads
me to think that the service was a loss making service designed just to
increase the customer base by exploiting network effects). Referring to
the Norwegian principle 2 above, it's not enough to receive and send
content and use applications of their choice; it needs to happen on a
level playing field. In the happy times when networks were separated,
voice came in nice and handy G711 quanta, which was handled on one
wire, while the fully liquid best effort traffic was handled on another
wire. The "interoperability" concept was an easy one: either fully
guaranteed or no guarantee. Interoperability is an on/off condition.
The merge of the voice traffic on IP networks prompted for some network
management, but it still is voice in handy well standardized quanta and
the good old on/off interoperability concept applies. What happens when
a new set of services appear which are neither latency-guaranteed voice
nor best effort ? there's not just 2 forms of traffic, there might be
many more, and interoperability is not just on/off but there might be
many colors of it.
Interoperating is not enough. You need to interoperate with a
comparable Quality of Service; otherwise there will be users with a
better users experience which would, in turn, lead to network effects.
Interoperability, without comparable QoS interoperability, paves the ground for anticompetitive behaviours.
(*)
Be it noted that, while there's a bitstream wholesale offer, any ISP
can compete anywhere in a country, purchasing wholesale at cost
oriented (and hence flat) conditions, so any operator can provide a
flat rate retail offer, so there will be one who does. As bitstream
wholesale offers are not part of the regulation in the wireless market,
so it is an excludable reource; furthermore spectrum is limited (rival)
so traffic discrimination based on pricing in wireless is likely to
stay with us for a long time, unless there's a very competitive
wireless market.
Quinta's Network Neutrality principles
Based
on the above, one could think that, rather than trying to give a
technical definition of Network Neutrality, it could be possible to
enumerate a set of principles which avoid the aforementioned problems,
to the benefit of the customer.
So, this is my proposal for five principles:
- users have the right to know what they are offered by their ISPs
- users have the right to purchase, based on their free choice, an
access which encompasses some network management; users have the right
to purchase a best effort access, with non discriminatory pricing, with
similar characteristics (**) (ref. B1)
- to ensure privacy of communications, traffic should not be discriminated on content (ref. A)
- traffic should not be discriminated by technology or pricing, based on who the endpoints of the communication are (ref. C1)
- operators offering a determined quality of service on their
network should provide the same quality of service at the peering
points to other operators who request it, at reasonable, non
discriminatory, transparent conditions. (ref. C2)
(**) e.g. best effort should not cost more than managed and if there's a 20Mbps managed, there should be a 20Mbps best effort
I'd welcome your comments...
