I attended a meeting of the ITU Joint Rapporteurs Group to Study Group 3. This was held in Brussels and lasted 2.5 days.
It concerns a proposal to change the pricing basis of IP connectivity on the Internet from “Sender Keep All” (where each party pays the cost of their own connectivity and keeps all the income from their networks’ customers) to one where there is a global accounting and settlement regime modelled on the PSTN “call-minutes” accounting and settlement.
It is also, on another (political) level, about how much influence the ITU should have on Internet governance.
I have attempted to write this report assuming no knowledge of ITU processes or the background to these proposals; subject to that constraint, I have attempted to give a brief write-up of the substance of the discussion.
This meeting is mainly relevent to (most to least affected):
* International carriers
* All those who buy international transit from the international carriers
* Anyone who buys or sells transit
Also, one side of the argument in this meeting specifically cites European IXPs as an alternative model to the proposals. It is possible that this will result in a direct request to Euro-IX for knowledge transfer or other assistence in establishing IXPs in the Developing World; see paragraph “IXPs: an alternative model”.
The “problem” addressed by this meeting is that international transit is very expensive for Developing and Least Developed Countries. The meeting considers how to modify the basis on which international connectivity is paid for so as to ensure the Developing World can take a full part in the “Information Society”.
Some LDCs report that the total international IP transit capacity out of their country is 1Mb/s or 2Mb/s.
Several years ago Telstra, the Australian incumbent operator, had to pay much more for international transit than US and European ISPs because there wasn’t as much international capacity to Australia. Telstra got together with the Chinese to propose that the ITU should set the framework for the pricing of international transit.
This work was done in ITU Study Group 3.
The proposal was controversial, and although LDCs liked it the USA was having none of it. Eventually “ITU Recommendation D50” was born which says, in very broad terms, that international internet connectivity should be charged on a basis agreed between the parties having regard to relevant factors. It is not at all prescriptive.
Even so, the USA (accompanied only by Greece) declined to sign on (“took a Reservation”, in the jargon), on the grounds that this is none of the ITU’s business.
Those countries that were very keen on D50 felt that it didn’t go far enough, and established two Rapporteur Groups reporting to Study Group 3:
- one to investigate ways to modify D50
- one to investigate traffic flow accounting for internet traffic
“Traffic flow accounting” is investigated with a view to mandated per-packet accounting and requiring financial settlement between networks based arithmatically on this accounting; all this is copied over from international PSTN settlements.
This meeting is a meeting of both Rapporteur Groups meeting jointly.
ITU structures and processes
ITU Recommendations, while formally having no authority to compel anyone, are in practice real regulatory instruments. Many countries simply incorporate ITU Recommendations wholesale into their domestic regulatory structure, as the framework that guides the exercise of the National Regulatory Authority’s discretion; this can then be interpreted and “compliance” with the local framework can be “enforced” using the usual tools, including fines and withdrawal of operating licences. In almost any country an ITU Recommendation will be given weight in the policy making process.
The Joint Rapporteurs Group, consists of both government representatives and representatives of private sector organisations that are ITU Sector members. Attending this meeting were:
* government representatives from USA, UK, Australia, China, Vietnam, India, Korea, Togo, Mali
* US based international carriers: AT&T, MCI, Sprint, Verizon
* BT, Telefonica, Telstra (Australia)
* incumbent operators from China, Vietnam, India, Korea, Togo, Mali
China is the chair of the “traffic flow methology” Rapporteurs Group and is highly partisan. That group has never met, and China simply advances its own position. This has caused criticism within Study Group 3. The chairman the other Rapporteurs group and of the joint meeting is from the European Commission; he is trying to stay neutral.
Private organisations are both permitted to attend and to speak in their own right at the Rapporteur’s meeting. By contrast, Study Group 3 consists of government representatives and their invitees, and private sector organisations do not speak unless invited to by their government.
The Rapporteur Groups simply report to Study Group 3 and do not have any decision-making authority.
Any decision is ultimately endorsed by and in the name of the “World Telecommunications Standardisation Assembly” (WTSA).
Governments may make their own proposals directly at Study Group 3 or even at WTSA, but in general the higher up the chain they introduce their proposal the less likely neutral countries are to support it.
The proposal that Recommendation D50 should be modified has two main constituencies:
* Less Developed Countries (LDCs) want some form of cost-sharing or “settlement fee” so that they don’t have to pay the full cost of building networks out to their remote areas.
* China (and sort-of Australia, see below), who believe that it is unfair that they have to pay for international connectivity as they have mature substantial local internet “markets”.
Both groups see peering as being “free Internet connectivity”, from which they feel excluded.
Apart from Australia there is no real division between the position of the national incumbent telco in each of the above and its government.
Against this are ranged:
* International carriers (AT&T, MCI, Sprint, Verizon, Cable & Wireless), who want to be able to recoup the cost of building out their networks to the third world by charging their customers there;
* US government, which supports business which are often US based, and which is also opposed in principle to the notion of the ITU extending its remit into Internet regulation in any form (US even took a “reservation” from the original D50);
* UK government (DTI) believes the US position is desirable but unrealistic, and wants to co-operate on a “harmless” modification to D50 so as to forestall further political pressure by the Chinese and others in the future;
* Telstra, the Australian incumbent, which has seen its own market position change substantially over the intervening period. It is now thoroughly and avowedly embarassed that it started all this off, and is trying to persuade its own government to reverse its support.
* European Commission chairman of the Rapporteurs group, who is trying to broker a compromise along the lines of the DTI position.
Proposals made at previous meeting
Proposal detail: China
* China’s proposal allowed individual contracting parties to agree terms, but sets a baseline recommendation for how to proceed where they cannot agree.
* China’s proposal was for “countries” with network traffic of between 20% and equal traffic to that of the “country” to which they are being connected to pay half-circuit costs for the international link capacity. Countries with a “traffic ratio of between 0.8 and 1.2” of each other should also peer their internet traffic with each other.
* China is citing research conducted by ETRI into IP accounting. This, it claims, shows that it would be possible to introduce an accounting and settlement model on the Internet that is similar to the PSTN settlement model.
* China had introduced a “study” to show the relative Internet traffic between “countries”, as measured by the ITU Bangkok office. The carriers produced detailed questions to investigate the methodology and origins of this data, but my view is that neither party was being entirely sincere. On the Chinese part, the data was probably simply an attempt to show that if their proposal were adopted then few countries would be affected (“so why not just go along with it”). On the carriers’ part they knew full well that the numbers were utterly meaningless and of highly questionable provenance, and that the Chinese couldn’t answer the questions even had their representatives had either the expertise or information available; their purpose was, I believe, simply a political ploy to brush the Chinese aside by saying “You submit this data in support of your proposal and we don’t understand the data so we cannot consider the proposal until we do understand it”.
Carriers’ view of Chinese proposal:
* It must be vigourously opposed.
* More specifically, they say that it is hard enough at the moment to make a business case for building out your IP network to anywhere where you aren’t already, and this would prevent any further expansion.
Proposal detail: EU chairman (supported by DTI)
* The Recommendation should be modified to say that when pricing international internet connectivity is priced there are lists (enumerated in the modified Recommendation) of technical issues that might be relevant to pricing;
* The Recommendation should still make clear that such negotiations are entirely voluntary and commercial, and it is not part of the Recommendation that the parties be obliged to consider any of those factors.
Carrier’s view of EU chairman’s proposal:
* If they trusted that the impact of modifying Recommendation D50 would be only what was said in this proposal then they would be largely content but
* They don’t like the idea of the ITU having a say, on principle; and
* This proposal is too detailed; and
* They believe that National Regulatory Authorities in LDCs will take the Recommendation and give it teeth by making compliance (including taking whatever the NRA considers to be “sufficient” account of each enumerated factor) a condition of licensing, thereby obviating the text in the modified D50 that declares these factors voluntary.
New industry proposal at this meeting
The international carriers got together and produced their own proposal for modifying D50. Their proposal is both very broad (rather than detailed) and permissive (rather than prescriptive). It is also to be attached as a “Supplement” to D50, which means that D50 itself will not actually be altered. A “Supplement” is an ITU document that is considered a reference but is not considered an integral part of a Recommendation and is not deemed necessary to interpret it. In short, it is the lowest grade of ITU document.
Nonetheless, this is the first time the international carriers have made a positive proposal, rather than simply making a case for not developing D50.
One American carrier spokesperson described their own (previous) position as “sitting down and chanting ‘Hell, no! We won’t go! Hell, no! We won’t go!’”
Reactions to industry proposal
This movement from the industry was a surprise to everyone else. The UK government immediately gave it its support, as did BT and Telefonica (which have previously sat fairly quietly while the Americans made the objections).
The Australian representative indicated that he was very inclined to support this too (perhaps this is seen as being a way out of their embarrassing divergence from Telstra), although he will have to consult before officially signing on. Nonetheless, Australia is still seeking further research into IP traffic flow accounting.
China is only interesting in advancing the traffic flow accounting research and in persuading others to accept that it is “proven” to work.
Traffic flow research
The meeting heard technical presentations from NEC in Germany and ETRI.
The NEC presentation stated that traffic flow accounting requires you to measure traffic flows by aggregating IP streams into a “flow” and analysing the contents of IP packets (not merely the IP header) so as to determine the application protocol being used (e.g. SMTP, HTTP, eDonkey, Quake etc). He claimed that the state of the art is capable of doing this fairly but not completely reliably at 10Gb/s. As such he deemed it suitable for access accounting but not for global internetwork accounting and settlement.
The ETRI presentation reported on research they had done to build equipment capable of the above. Using a standard Intel-based PC of moderate spec and some custom hardware (a PCI card they built themselves) they were able to achieve detailed application-level flow reports in the laboratory at the following speeds:
* OC3: tested and working now
* OC12: passed “usual case” test, in testing for “worst case” test
* OC48: expecting to test in Q4 2004
They claim to have tested on a Korean IXP too; CERNnet refused them permission to connect this prototype equipment to their network when asked.
It was admitted that application-level content inspection of IP packets is not possible when the content is encrypted.
Various other problems were cited by industry representatives, including assymetric routing, data protection issues, and the implications of “best efforts” packet delivery in an environment with PSTN style accounting and settlement.
IXPs: an alternative model for promoting global access
Rather than change the pricing and payment model of the Internet, international carriers are keen to cite alternative ways developing countries can gain cheap fast international internet connectivity. These include:
* a liberal pro-competitive regulatory environment
* building local IXPs to keep local traffic local
both of these measures are advanced as a means of increasing teledensity and providing an environment where there is a commercial incentive to build additional capacity.
Europe is cited as the proof: Europe has followed this path, and “as a result” it has high-capacity, cheap connectivity.
The international carriers would like to see new IXPs built for parts of Asia-Pacific and sub-Saharan Africa using European knowhow. Connection to these “regional hubs” from individual countries would possibly be subsidised by the World Bank out of development funding. This would hopefully provide sufficient teledensity that there was a truly competitive commercial incentive to sell wholesale transit into the hub at something closer to European or US transit rates. ITU involvement in this proposal may not be sought, but it would be advanced as a reason to say that the ITU did not need to progress with creating a new accounting and settlement regime.
The Joint Rapporteurs Group will report back to Study Group 3 recommending
- further research on traffic flow accounting
- adding the industry’s proposal as a “Supplement” to D50
It will also report the discussion around existing traffic flow accounting research by ETRI and NEC that occurred at the meeting. There was no consensus. The Group will also recommend that research into traffic flow accounting be expanded to include “economic, cost/benefit, legal and security” issues, as well as purely technical ones.
Study Group 3 next meets 31st May-4th June 2004.
WTSA next meets 5th-14th October 2004.
Whether the Rapporteur Groups meet again depends on the outcome of the Study Group 3 meeting.