Paris, 7 October 2015
Government’s ability to solicit advice from ARCEP
The Act on Growth, business and equal economic opportunity that came into effect on 6 August of this year brought significant changes to Article L 32-1 of the French Postal and Electronic Communications Code (CPCE), and gives Ministers responsible for electronic communications and postal affairs the possibility of soliciting advice from ARCEP on all matters that fall under its purview. If parliamentary committees were already able to solicit advice from ARCEP on matters within its purview, in the past the government was only able to consult with ARCEP by soliciting the Authority’s advice or opinion on a specific bill or piece of draft regulation.
Emmanuel Macron, France’s Minister of the Economy, Industry and Digital Affairs, thus solicited ARCEP’s advice on the current structure of the use being made of bandwidth on internet access networks in France, and on the possibility of measuring it (opinion issued on 7 July, which is the subject of this press release). This first request for advice was made in during the planning stage for the new aforementioned framework, which the Act brought into effect. Additional requests for advice have been made since then.
By and large, these independent analyses and expertise are intended to be made public, with any elements protected by trade secrecy removed when applicable. Furthermore, this advice can pertain to specific technical points and serve to shed light on government works in progress, and ARCEP’s Executive Board wanted to give the parties requesting the advice a three-month period to draw their own conclusions before publishing it (unless requested otherwise by said parties).
ARCEP opinion on the structure of the use being made of bandwidth on internet access networks in France
It is within the context described above that, today, ARCEP is publishing its opinion of 7 July 2015 on the current structure of the use being made of bandwidth on internet access networks in France, and on the possibility and means of collecting information that would enable detailed knowledge of the nature and source of these traffic streams. This contribution to Government debates does not concern the future use that could be made of the findings of these measurements.
The conclusions of the opinion are as follows
The task of measuring internet traffic is complicated by the diversity of the methods used to relay this traffic. IP packets may indeed take a variety of different, more or less direct paths to reach the end user, as part of a single exchange. These paths travel over multiple networks and can rely on cache servers – where a version of the requested content is stored – that are located more or less close to the end user.
Moreover, there is no single, unequivocal, infallible or exhaustive method for associating a traffic stream with a service. The traffic being relayed is indeed multifarious – e.g. live and on-demand video, telephone calls, files, websites, etc. – and one characteristic feature of the internet is the strong decorrelation between content, services and applications on the one hand and, on the other, the networks used to relay them in a way that is largely undifferentiated. Identifying the type of service being relayed would therefor require a very detailed analysis of the streams which could, in addition to raising concerns in the area of data privacy, be rendered inoperable by the encryption of the data being transmitted.
These limitations are intrinsic to how the internet operates, but also constitute its strength since they ensure both its reliance and its openness.
It was with all this in mind that ARCEP assessed the different options for measuring the use made of bandwidth one the internet. These options were evaluated based on the following criteria: technical feasibility, the cost and complexity of implementation, the accuracy and verifiability of the collected data, while complying with the principles of net neutrality and confidentiality of correspondence.
In terms of level of analysis, the Authority concludes that measuring the volume of data traffic on a given link thanks to SNMP would make it possible to obtain accurate and verifiable results, and be relatively easy to implement, compared to other forms of analysis such as Netflow/IPFix or Deep packet inspection (DPI).
As to the point in the network where the measurements should be taken, ARCEP concludes that applying this method (i.e. measuring volume using SNMP) at internet exchange points would prove a reasonable undertaking for internet service providers (ISP) from a technical and economic standpoint.
This type of measurement would have two limitations, however:
· first, because it involves merely counting the number of packets transiting over the network, the stream is identified as coming from the interconnection/peering partner. In most instances, then, this would be a technical intermediary (transit operator, Content delivery network, Internet Exchange Point (IXP) manager, etc.) and not a service provider;
· second, because measurements are performed at interconnection points, they do not make it possible to take into account (in any event not directly) the volumes of traffic being relayed under special conditions such as multicasting, peer-to-peer exchanges between two subscribers with the same ISP, and transmissions from an ISP’s own web hosting centre or a server on its own network.
As a result, there is also a risk that implementing such a measurement system would lead certain players to adopt bypass techniques. Except for peer-to-peer traffic which could not be associated with a service.
Paris, 7 October 2015