COMPUTERS & LAW · up a carrier's fixed assets without yielding incremental revenues; and...

COMPUTERS & LAWJOURNAL FOR THE AUSTRALIAN AND NEW ZEALAND SOCIETIE5,

FOR COMPUTERS AND THE LAW

Print Fbst— PP233867/TD008

Editors: Brendan Scott, Simon Pollard and Kent Davey ISSN 08117225Number 38 April 1999

Internet Connectivity: Open Competition in the face of Commercial Expansion

Bernadette Jew, Rob Nicholls & Michael Reede, Gilbert & Tobin

1. INTRODUCTION1.1 The In tern et could be

considered a carrier's worst nightmare or, perhaps, a dream come true. It poses a potential competitive threat for every provider of telephony, broadcasting and data communications services, since it is substitutable for all existing m edia. At the same tim e, internet-related businesses are substantial custom ers of existing telephony,broadcasting and data companies.

1.2 The shift from circuit- switched to packet-switched netw orks requires a new approach to interconnection - not just because the internet is fundamentally different from other com m unications technologies, but because it presents new cost structures and revenue opportunities. For exam ple, the in tern et exploits every weakness of the traditional telephony carrier: motivating end users to stay logged on all day, and tying up a carrier's fixed assets

without yielding incremental revenues; and substituting e- mail for potential long-distance voice traffic th at would otherwise pay for the carrier's fixed costs of infrastructure. From a revenue perspective, however, the internet not only creates alternate distribution channels for pre-existing content but, more importantly, it permits the delivery of new and hybrid forms of content - w ith the div erg en t values attributed to various kinds of content presenting yet another

Continued on page 3

in this issue...

Internet Connectivity: Open Competition in theface of Commercial Expansion ................1by Bernadette Jew, Rob Nicholls & M ichael Reede

From the Editors* D esk..... ................ 2

Telecommunications Carrier Licensing ......... 20by Kent Davey

Domain name disputes: A view from the Antipodes......... ...................... 25

by Stephen Lance

Linux for the Legal Community: A Primer.... 34 byConZymaris

The Dawn of a New Dark Age Censorship and amendments to the Broadcasting Services Act.......................................................... .......... . 39

by Brendan Scott

From the Minister's Office........ ............ 42

From the Editors' Desk

Welcome to our first issue for 1999 focusing on the rapidly growing area of telecommunications. On 1 July 1997 a new telecommunications regulatory fram ework was introduced which was designed to achieve full and open competition in the provision of b oth telecom m unicationsinfrastructure and services.

Since 1 July 1997 there have been m any new entrants to the telecommunications industry with the removal of regulatory barriers to en try including the limit on the number of licensed carriers. At last count there were some 28 carriers which had been issued licences by the A ustralian C om m unications Authority.

The new regulatory framework was intended to bring the regulation of com petition w ithin the telecommunications industry more closely into line with general trade practices law. It provides the A ustralian C om petition and Consumer Commission (ACCC) with the pow er to issue a so-called 'competition notice' to deal with anticompetitive conduct in the industry. To date the ACCC has had occasion to issue several such notices dealing with Internet access and customer churn.

In our lead article, Bernadette Jew and Rob Nicholls investigate internet interconnection within Australia and the effects of the ACCC's competition notice against Telstra last year. As the

authors discuss, prior to the issue of that notice, Telstra had refused to enter into interconnection agreements with any other ISPs w ithin Australia (including Optus) despite the fact that equivalent traffic flows and network services could be proven. Their paper

in part discusses the effects of that notice, some of the issues faced in d eterm in in g app ro p riateinterconnection arrangements, and some risks involved in misunderstanding the market.

In the next paper, Kent Davey discusses the licensing of telecommunications carriers under the new regulatory framework. Kent looks at the process of applying for a carrier licence and examines the key questions which are relevant to determ ining w hether a person is required to hold a carrier licence. Following Kent, Stephen Lance gives us a much needed refresher on the law of domain names. The last 18 m onths have seen a lot of activity discussing proposed domain name regulation schemes. Fundamental to that activity has been the tension between domain name seekers and trade mark holders and, in particular, the privileged position some trade mark holders consider themselves entitled to. Stephen presents a view from dow n un d er on these contentious issues.

O ur last paper, by Con Zym aris, discusses the only operating system Microsoft sees as a threat - partly

because it's free. Linux, the creation of pooled programming effort from around the world, has increasingly been making headlines in IT press and is seen as a viable alternative to mainstream operating systems. In his column Con provides an insight into what Linux is, with some answers for a firm's IT manager. Definitely one worth watching.

Last year we broke with tradition in publishing the president's report of the Victorian Society early (that is, not in the first edition of the new year). This year, we're breaking the tradition again, moving reports to the second journal of the year. The editors would like to th an k everyone for their positive feedback about the journal. We'd also like to remind you not to be shy with your articles. If you have som ething of in terest to your colleagues, please send it in.

Legislatively, the good news is that we now have a Commonwealth Year 2000 Disclosure Act as well as the Electronic Transactions Bill (but at least a step in the right direction).

Finally, what would the editorial be w ith o u t a short polem ic about something? The lay down misere for this edition would have to be the Federal G overnm ent's recent censorship initiatives for the internet. To find out why everyone loses, see our article at the end of this edition.

Government ReleasesINTERNET CENSORSHIP BILL see Brendan Scott's article on page 39

2 COMPUTERS & LAW

Internet Connectivity

Continued from page 1

challenge to the traditional 1.4 interconnection model.

1.3 The maturing of the Internet and the dynamics of Internet in terco n n ectio n are quite d ifferent to the dynam ics en co u n tered in other in d u stries th at have been established according to normal commercial structures.The culture and structure that defined the Internet's early developm ent are still major influences on the industry. M any of the cu rren t contentious issues relating to Internet interconnection arise out of principles and solutions which were acceptable for a nascent structure supporting non-commercial applications, bu t w hich are no longer ap p ro p riate in a m ature commercial industry

Internet interconnection is currently in a state of turmoil throughout the world. Internet backbone providers are no longer p repared to rem ain b o u n d by In tern etinterconnection arrangements which fail to take adequate account of associatedinfrastructure costs - costs which are grow ingexponentially w ith the increased dem and for b a n d w i d t h - h u n g r y applications. At the same time, aggrieved In te rn et service providers (ISPs) have now come to expect free access to the Internet backbone,1 and are claiming that the refusal of Internet backbone providers to make available settlement-free peering to ISPs on an ongoing basis is "discriminatory".2

1.5 As a consequence, the rules for Internet interconnection are still in the process of being w orked out - and in this uncertain environm ent, Internet interconnection is an inherently risky business. The new technology presents its own interconnectioncomplexities, with m ultiple layers of virtual networks built one over the other - so that an operator at any layer of the infrastructure will have its costs determined by the prices charged by the virtual network below it, while its prices, in turn, will determine the cost structure of the layer above. F urtherm ore, w hilecommercial principles would suggest that m oney should flow towards those operators which produce value, this does not always happen in practice

COMPUTERS & LA W

Brendan Scottc/» Gilberts Tobin SO Carrington Street Sydney 2000Australia Tef+ei2saa?aoa4Fax+612 9367 3111 e-mail: bsco8@^aw.com ,au

Editors

Simon Pollard cACifbertS Tobin 50 Carrington Street Sydney 2000 Australia Tel+612 036? 310$Fax+612 9667 6111 e-maifcspoliard^tlawcom.au

Kent Daveyc/-Austraiian Communications AuthorityP O Box 13112, taw Courts MelbourneVIC 6001 Tel +61 3 9963 6049 F a x +61 309636609 e^maltkentdavey^aca.gov.au

Subscription: $32.00 per 4 Issues

Advertisements: Inserts $300.00: for advertisements w ip n t ie journal, rates and information will be provided by the Editors on request

Articles, news items, books for review and other items of interest may be sent to the Editors. Journal contents may be reproduced if the source is acknowledged.

COMPUTERS & LAW 3


(a hangover of the historical expectation that Internet access should be free for all). The uncertain ties of the environment are aggravated by the fact th at th ere is currently no consensus on the issue of how to attribute "value" to the various elements of Internet interconnection.

1.6 The industry is still working through the basic, and yet daunting, task of determining the optim al set of Internet i n t e r c o n n e c t i o n arran g em en ts. As this dev elo p m en t unfolds, regulatory authorities are faced with the question of whether an unregulated m arket can provide a "socially optimal" level of interconnectivity at an acceptable cost - and if not, what forms of regulation are required? An add itio n al element in the equation is the burgeoning growth of vertical in teg ratio n and vertical p artn ersh ip s, w hich create further potential for regulatory intervention in an industry which has experienced spectacular grow th and innovation to date without the constraints of regulation.

1.7 While carriers have been busilyw orking out how to re dim ension their Public Sw itched TelephoneNetworks (PSTN), Internet protocol telephony has grown to the extent that it looks likely to replace the PSTN in many situations. Accordingly, In tern et in terconnection presents one of the m ajor challenges for the telecommunications industry over the next ten years - and the industry's response to that challenge will be crucial to its future com petitiveness and growth.

2. THE NEED FORINTERCONNECTION

2.1 At the heart of all networks is interconnection - the ability of one entity to connect to other

entities. This is particularly so in relation to the Internet, where interconnection is the glue that holds the "network of netw orks" together. The objective of global connectivity has not always been a feature of the Internet - initially, the sole purpose was the sharing of rem ote computer resources among as many sites as possible. Global connectivity only became a possibility as ever-increasing num bers of individual netw orks connected to the Internet. However, we have now reached the stage where each end user connected to the Internet has the expectation of global connectivity to all other end users connected to the Internet.

2.2 There are many alternativeinterconnection arrangements in the layered structure of the Internet that can result in full connectivity among end users - encompassing a mix of direct connections, transit traffic, public exchanges, private exchanges, "m ulti-hom ing" and international connectivity. Full connectivity does not require all netw orks to establish businessrelationships w ith one another, but rather that the m atrix of interconnections th ro u g h o u t the in dustry enable all end-users to reach one other through seamless connectivity am ong the millions of routers, services and users.

2.3 The optim al degree ofinterconnection can only be achieved if those interconnection arrangements do in fact reflect the actual costs associated withi n t e r c o n n e c t i o n . Interconnection charges which bear no relationship to the actual costs associated with interconnection willinevitably create distortions in the m arket, w ith potential risks ranging from heightened barriers to entry (and the

lessening of competition) to the undermining of incentives for investm ent ininfrastructure.

2.4 This p ap er canvasses the practical, economic and regulatory factors to be taken into account as the Internet in d u stry works tow ards identifying and achieving that optim al degree ofinterconnection.

3. BACKGROUND TO CURRENT INTERNET DEBATE

3.1 It is essential to view the In te rn et interconnection debate in the context of the evolution of the Internet - from an academic, non-commercial facility to a commercial tool characterized bv exponential growth. The Internet was not designed with commerce in m ind, but simply began to assume a commercial role once com m ercial organizations recognized the enorm ous benefits that it could deliver:

• The growth of the Internet was originally driven by the goal of Internet engineers to make the n etw o rk function as widely and as quickly as possible, with the limited objective of sharing remote computer resources among as many sites as possible.

• Before the Internet "went com m ercial", it m erely presented an opportunity to sell leased lines to academ ic institutions. Applications by end-users w ere largely n o n commercial, and were more in the nature of gadgetry than the part-lifestyle, part- commercial applications that characterize it today.

• As the In tern et became more widely available and used by com panies and in d iv id u als, the prim e o p p o rtu n ity for carriers became the provision of

4 COMPUTERS & LAW


b an d w id th , w ith little indication of the commercial explosion that was to follow.

Com m ercialization of the Internet: changes in usage, utilization and infrastructure requirements

3.2 It was not until about 1993 thatthe Internet began to move from the non-com m ercial model towards the Internet model that we know today. The num ber of users grew exponentially as more and more people started to notice the attractiveness of w hat looked like, if not a free lunch, then something close to it. This was accompanied by dramatic increases in capacity utilization levels due to the type and volum e of data accessed (reflecting the increasing dem and for b an d w id th h u n g ryapplications), as well as an increase in the average capacity requirem ents of customers (who were simply spending far more time on the Internet). Furtherm ore, the World Wide Web created a split between the end users (who im port data) and content com panies such as Web hosting firms (who are data exporters), resulting in major traffic imbalances.

3.3 As a result, the costs of Internet backbone providers increased, and are continuing to do so, not only due to the need for upgrading capacity, but also as a result of the shortening of capital investm ent cycles in network upgrades - from 24 m onths to periods approaching 6 m onths. Internet backbone providers are now facing the need to find ways of increasing their infrastructure, while at the same time their more profitable base of private netw orks is beginning to erode as the corporate world exploits the cost advantages inherent in hybrid internet applications such as intranets.

C urrent in terco n n ectio n "problems" once provided the perfect solution

3.4 The failure of many Internet interconnection arrangements to take adequate account of infrastructure costs is therefore a reflection of the historical perception that the Internet is "free". It is also a reflection of the fact that Internet usage, utilization p a tte rn s and infrastructure demands were quite different during the early stages of In tern et development. In the early days, Internet backbone providers and ISPs generally swapped traffic freely - since early Internet applications like file transfer protocol led to more or less sym m etrical traffic am ongst them . The capital costs of infrastructure were comparatively moderate, with there being an adequate am ount of "slack" in the average usage of each bit pipe. F urtherm ore, no-one was trying to make a profit.

3.5 In addition, routing practices which are problematic today were perfectly acceptable in the past. For exam ple, the technique known as "shortest exit ro u tin g " (som etim es described as the game of "hot p o tato "3) was perfectly appropriate when traffic flows were comparatively low, but creates problems in the current environment:

• Shortest exit routing means th at data goes onto the receiver's network at the earliest p oint. This approach saves the sender ISPs the costs associated with bandwidth, since they are able to reduce the bandwidth costs associated with hauling data to the p o in t closest to the customer, and instead pass that burden to the Internet backbone provider at the earliest point. This practice may even involve an ISP linking one of its own subscribers to another of its

subscribers via a backbone provider, rather than directly via its own links.

• Shortest exit routing works to the advantage of an ISP where it connects with a backbone provider on an sender keeps all (SKA) peering basis and traffic flows are asym m etrical (with the ISP sending more traffic than it receives). Because the ISP is not required to pay for traffic sent via the backbone netw ork, it has every incentive to offload data onto the Internet backbone at the earliest possible point - without being required to carry an equivalent volume of the Internet backbone provider's traffic via its own network.

• By com parison, the dynam ics are quite different w here an ISP connects to the backbone provider u n d er a settlements arrangement, as the ISP has an incentive to retain data on its own netw ork for as long as possible w here the costs associated with off-loading that traffic onto the Internet backbone p ro v id er's netw ork are higher in comparison.

• When networks were all roughly the same size, the game of "hot potato" was treated as a simple reality of In tern et provision. However, traffic flows and volum es have changed over time so that there is now both a traffic and cost im balance betw een netw orks th at supply connectivity to content sites (and are net exporters of data) and those networks with many subscribers that request the data (which flows over their own provider's network). These im balances have now reached such a scale that shortest exit routing is no

COMPUTERS & LAW 5


longer ap p ro p riate . It imposes severe capacity demands and burdens on the perform ancecapabilities of In tern et backbone providers, having a significant effect on the ability of backbone providers to recover the costs of their investment in infrastructure.

3.6 In summary, it is not so long ago th a t m any of to d ay 's "problems" provided a viable solution w ithin the limited param eters of the n o n commercial Internet - with the added attraction of being easy to implement. However, usage of the In tern et has subsequently altered so drastically that such practices are no longer sustainable in a commercial environment. As a result, we are now on a collision course between the spiralling b an d w id thconsumption of users, and the expectation on the part of ISPs of an en titlem ent to free Internet interconnection.

4. THE DISTINGUISHING FEATURES OF THE INTERNET

4.1 In any consideration ofInternet interconnection, it is essential to identify the distinguishing aspects of the Internet - given that many of the assumptions and business models that have been applied to trad itio n alt e l e c o m m u n i c a t i o n s interconnection arrangements will fail to deliver viable com m ercial results on the Internet.

Absence of end-to-end circuits4.2 As noted at the outset, the

Internet is based on packetswitching, with there being no dedicated connection - the data is broken up into packets which may be delivered via totally different routes over the Internet, and reassembled at their ultim ate destin atio n .

There is no lim it as to the number of networks which a packet of data may transit before it reaches its destination. This is in stark contrast to the trad itio n al telep h o n y netw orks, which switch or assign a dedicated end-to-end circuit for every call - and are able to charge on the basis of minute-by-minute and circuit- by-circuit.

US-centric industry4.3 A nother key feature of the

*

Internet is that it is US-centric. This is partly a reflection of the fact that the Internet initially started as a national US network, and partly because the US is the source of the world's most popular Internet content, and the pre-eminent cultural influence as globalization becomes a reality. Today, half of the Internet's users and 58% of all content hosts are based in the United States.4 While this situation will inevitably change over tim e, non-US backbone providers are currently exhibiting some impatience w ith the costs entailed in acquiring substantialinternational capacity to and from the US. Non-US networks were initially happy to pay for both ends of the transoceanic circuits required to connect to US backbones, but now claim that this practice requires them to effectively subsidize US Internet users (to the extent that US users are using those same links to access content outside of the US). Allegations of discrimination have been referred to the International Telecommunications Union, amongst other international bodies.

Layered nature o f the Internetindustry4.4 In the context of

interconnection, a key feature of the Internet is the layered nature of the Internet industry - with the "backbone cloud"

comprising a series of Internet backbone netw orks which provide the base layer of the infrastructure. The layer in w hich an entity operates w ithin the in d u stry will inevitably determ ine th at entity's position within the Internet food chain. The lower down in the chain, the more an entity will find that those above can raise prices or do other things that will affect its cost of operation.

4.5 The layers of the In tern et in d u stry can be described generically as follows:

• An Internet backbone is generally a national backbone (or in the US, it may one of the four super- regional backbones, Cable & Wireless, Sprint, UUNET or AGIS, which between them account for at least two-thirds of the backbone connections in the co u n try )5. Each Internet backbone is an overarching network to which m ultiple natio n al and regional networks connect, and which generally does not service directly any local networks or end users.

• There will often be a seriesof networks "hanging off" each Internet backbone (ie: connected to the backbone by way of fibre optic cable). These netw orks aregenerally operated by large ISPs utilizing the Internet backbone infrastructure constructed by thet e l e c o m m u n i c a t i o n s carriers. These large ISPs are able to interconnect with each other and direct traffic betw een their respective networks, with traffic only being directed over an Internet backbone where it is necessary to access an o th er In tern et backbone or any network linked to another Internet backbone. W hile these large ISPs have a national

6 COMPUTERS & LAW


presence, they are not treated as full peers to the In tern et backboneproviders.

• The large ISPs provideInternet access in turn to the regional ISPs (which construct regionalnetworks through leased lines).

• The regional ISPs provide Internet access in turn to those ISPs which have no investment in leased lines, and w hich simply resell Internet access direct to the end user.

• Online service providersalso utilize this infrastru ctu re for the purpose of supplying value-added services. They do not derive revenues by reselling netw orktransmission services, but rath er by offering proprietary content and specialized com m erce, selling advertisements, and providing users with the ease-of-use which comes from special softw are interfaces and custom er hot-lines.

4.6 At the retail level, Internet access is supplied as a bundled service in the form of online services together with such other additional services as:

• In tern et connectivity services, including support of various connections to customers (PSTN, ISDN, El, Microwave etc);

• online functional products including, in particular, email functions, browser software and other software products;

• caching for the temporary- holding of accessed web pages; and

• technical su p p o rt and billing functions.

4.7 The industry is not settled on what constitutes a full Internet backbone provider, as

compared with a large ISP and other layers or classes. This is a consequence of the fact that the dynamics of the industry are continually changing as networks mature at a rapid rate. Furthermore, Internet backbone providers can, and frequently do, choose to acquire the skills needed to offer Internet provider services, thus becom ing ISPs themselves. However, for the purposes of this paper, the reference to an In tern et backbone provider indicates an entity in its capacity as the supplier of In tern et infrastructure at the backbone layer of the industry.

Importance of the backbone layer4.8 All industry participants are

beholden to the In tern et backbone providers - it is by virtue of the quality and scale of their infrastructure that all other interconnecting parties prosper. By way of example, if Carrier A (a small regional ISP) interconnects with Carrier B (a large ISP) which in turn relies on the infrastructure provided by Carrier C (a backbone provider), the upper limit of Carrier A's service quality will be determ ined by the quality of Carrier C's infrastructure. Thus the backbone provider determ ines the quality of carriage services ultim ately delivered to end users. Indeed, it is the quality of infrastructure th at has attracted ISPs to interconnect in droves w ith Qwest in the USA.

4.9 Because of the significance of the backbone layer, the continuing growth and innovation of the Internet industry is dependant on the age-old dilem m a of achieving the ap p ro p riate balance between:

• encouraging competition at the backbone layer in order to provide the impetus for backbone providers to deliver innovative and price- competitive Internet services, and so facilitating greater connectivity and increased

competition at the ISP layer; and

• retaining incentives for ongoing investm ent in the Internet backbone infrastructure, which is crucial to the ability of the in d u stry to m eet the exponential increase in demand for bandwidth- hungry applications.6

Relevance of industry layers tointerconnection4.10 The layer in which an entity

operates is also relevant to the issue of an entity's ability to interconnect with other networks, ie: the criteria for interconnection turn on the extent to which a network is able to provide equivalent n etw ork services on a reciprocal basis to another netw ork. A netw ork operating somewhere down the chain is unlikely to be able to provide services of equivalent value to an entity operating at the backbone layer.

4.11 Wanting to have some degree of direct control over their own fate, many ISPs have tried to connect at the Internet backbone layer of the Internet hierarchy. This is one of the reasons why there has been such a heated response to the recent moves by five of the largest backbone providers in the United States to move away from public exchange points to a series of private interconnects (and so place them selves in a new hierarchy above the 40 to 50 ISPs at the public exchanges)7. ISPs are concerned about the additional costs incurred as a result of the shift away from SKA peering to settlements i n t e r c o n n e c t i o n arrangem ents. They also resent the fact that such a move results in their rejection from the backbone layer of the industry, so that

COMPUTERS & LAW 7


they are delegated to the status of "customer" - a fully-fledged dow nstream reseller of the Internet backbone provider - with possible flow-on effects in th eir interconnectionarrangem ents w ith other Internet backbone providers.

5. A COLLISION COURSE: PEERING V INTERNET EXPLOSION

5.1 There will inevitably be tensions betw een the economic incentives driving the various ind u stry p articip an ts in their interconnection dealings. For example, an Internet backbone provider will be looking for a retu rn on investm ent in bandwidth and other network infrastructure. By comparison, ISPs will instead be seeking to minimize their increm ental costs of transm ission by utilizing the infrastructure pro v id ed at the In tern et backbone layer to best advantage.

SKA peering or no-settlementpeering - the status quo5.2 Up un til recently,

interconnection on the basis of SKA peering or no-settlement peering was the accepted status quo arrangement for Internet interconnection - largely due to the non-com mercial and hom ogenous genesis of the In tern et. While the term "peering" is sometimes used generically to refer to Internet interco n n ectio n , the term refers m ore correctly to an interconnection arrangement with no financial settlement whereby:

• no m oney exchanges hands regardless of the volume of traffic or level of connectivity exchanged am ong providers - in contrast to the voice telephony business, which provides a well-established system of settlements;

• traffic is effectively exchanged under a barter arran g em en t - each network agrees not to bill the other network on the u n d e rsta n d in g th at the value of the netw ork services which the other n etw o rk is w illing to supply equates to the am ount which that network would otherwise have billed;

• while different bits of the message are likely to travel by different routes, the only cash payments are from the end-user to the ISP (usually in the form of a fixed m onthly fee) and from a particular ISP to the nearest high capacity backbone node (normally in the form of a capacity rental payment); and

• because each netw ork collects and retains all fees derived from its dow nstream subscribers and customers, this model is also referred to as a "sender keep all" or a "bill and keep" arrangement.

SKA peering - simply a hilling arrangem ent5.3 How ever, while peering is

simply intended to provide a m eans of sim plifying the billing adm inistration in circum stances w herenetw orks are providing equivalent services on a reciprocal basis, peering has in many instances been adopted blindly - either because of p reced en t, technological difficulties in implem enting any kind of alternative billing and payment arrangements, or possibly due to regulatory intervention (as has been the case in Australia). As a consequence, peering has come to be seen as an end in itself and to drive the economic imperatives, rather than simply providing an administrative outcome for a certain kind of interconnection arrangement.

SKA peering - the need fo requivalence5.4 Peering only makes sense only

where the costs and benefits of allowing reciprocally "free" access are equal. An SKA peering interconnection arrangement contemplates that the "peers" will be of roughly the same size—recalling the traditional sense of the word. Brock has argued that there are two conditions necessary for SKA peering to be viable. These are:8:

• that the traffic flows should be roughly balanced betw een interconnecting networks; and

• that the cost of terminating traffic should be low in relation to the costs of measuring and billing for traffic.

5.5 In practice, the criteria applied have in many instances been far broader than Brock suggests. The notion of "sim ilar size" has been determined not only by traffic flows and backbone capacity but also by num bers of customers, the services and content able to be accessed via a particular netw ork, the geographic reach of connectivity, and various other factors which networks have chosen to recognize as a basis for the free exchange of traffic.

Consequences of blindly adopting SKA peering5.6 Inevitably there will be some

subjectivity in the decision bv an Internet backbone provider as to whether or not to peer w ith an o th er entity. For example, greater geographic coverage maybe a determining factor which results in one netw ork being prepared to value the benefits derived from interconnection with another network as being equivalent. Similarly, an ISP with a very large client population within a limited geographic locality may provide significant value to another network.

8 COMPUTERS & LAW


5.7 F u ndam ental economicproblem s have arisen as a result of the adoption of peering interconnection arrangem ents betw eennetw orks which have disparate capital investment and w hich offer disparate value to each other. Where SKA peering has been adopted between networks which are not "peers" in the true sense of the word:

• it has resulted in costs being incurred by the Internet backbone provider, since the smaller player which is not investing in building national or international infrastructure inevitably derives greater benefit;

• ISPs with no investment in the building of national infrastructures and which are unable to offer eq u iv alen t value to backbone providers have come to rely on an "entitlement" to peering - thus ensuring that their traffic is transported across the global Internet at no cost other than the coordination costs to arrange i n t e r c o n n e c t i o n agreements; and

• it has meant that Internet backbone providers are not always fairly compensated even if the amount of traffic exchanged is equal - given th at In tern et backbone p roviders are forced to carry traffic further along their networks than the ISP (d u e to ISPs using the shortest exit routing or "hot potato" routing referred to in paragraph 3.5 above).

Im pact of SKA peering on long term investment in infrastructure5.8 As w ith mobile

communications, this skewed m odel of connectivity has allow ed the n etw ork to develop rapidly. H ow ever SKA peering provides no real platform for organic growth: revenues are required in order

to finance the large network upgrades an d increased capacity from which all networks derive benefit - be they direct downstream ISPs, customers or peers. Firms that have com m itted to m ajor in frastru ctu re investm ents have little to gain from interconnecting with smaller netw orks unless they are adequately compensated for the network services provided.

5.9 ISPs w ho seek peering arran g em en ts w ith large backbone providers may benefit in the short term, since it enables them to avoid any settlement charges. However, w h en peerin g provides a disincentive for the backbone pro v id er to build infrastructure, and ISPs are relying on that infrastructure for their Internet access, then the long-term benefits to ISPs and their customers are put at risk - the quality of network services will inevitably suffer if there are no incentives in place for the backbone pro v id er to continually u p g rad e and expand its infrastructure.

5.10 As argued by Farnon & Huddle,9 continued growth of the Internet depends on the development of an alternative to the SKA peering model. C om panies which have com m itted to m ajor infrastructure investment are increasingly reluctant to in terco n n ect w ith smaller netw orks because of the inevitable inefficiencies and burdens on their backbone performance arising from the SKA peering arrangements.

Analogy with traditionaltelecomm un icationsinterconnection principles5.11 The principle that a network

offering Internet backbone in frastru ctu re should be entitled to charge a higher price for interconnection than an ISP, reflecting the value of network services offered, is no

different to that adopted in trad itio n al telep h o n y interconnection arrangements. In telecom m unications we have seen the use of reciprocal arrangem ents in respect of terminating calls in order to eliminate differences between internal and cross-netw ork prices and to m itigate the advantages of m onopoly. However, the relative network size is still taken into account in determining the structure of component prices. A relatively larger netw ork still charges more than a smaller one for outgoing calls under such a reciprocal arrangement, due in part to the fact that origination fees increase as the originating network gets larger - because a large network, given its size advantage, commands a higher percentage of the revenue from outgoing calls.10

6. US DEVELOPMENTS: BREAK AWAY FROM PEERING

6.1 Approximately three years ago the big five Internet backbone providers in the United States established criteria for SKA peering. Since then, between 15 and 20 national ISPs have met the criteria - far more than ever imagined possible. This has placed a strain on the peering relationships between US networks, particularly as the traffic of the large Internet backbone providers has become considerably larger than that of their counterparts or "peers".

6.2 In the face of the emergingcollision course betw een peering and the ongoing requirement for expansion in in frastru ctu re , some US Internet backbone providers took the drastic step in 1997 of refusing to renew many of their previous peeringarrangem ents. Theytran sitio n ed th eir peering arrangements away from the large public exchange points in favour of setting up direct

COMPUTERS & LAW 9


interconnections with other networks - allowing improved levels of service in the technical aspects of interconnection.

6.3 As a resu lt of these developments, peering grew more prevalent as between ISPs themselves (both nationally and on a regional basis), and was boosted by the creation of m ore local n etw o rk access points - a logical outcome of the incentives which ISPs now face in developing new efficiencies, since they can no longer rely on free access to the Internet.

Chain o f events6.4 AGIS was the first Internet

backbone provider to unilaterally terminate peering arrangem ents in 1996. MCI subsequently term inated its connection to the C entral Internet Exchange (CIX) in February 1997, ending peering to several ISPs. In the m eantim e, Sprint began a policy by w hich its peers would have to pay in declining am ounts, according to the num ber of exchange points w here they conducted peering.

6.5 How ever, the case w hich received the most attention was that of UUNET, a subsidiary of WorldCom (and perhaps the largest In te rn et backbone provider in the world), which declined to renew peering agreements with a number of ISPs as from May 1997. UUNET stated that it would no longer accept peering requests from other netw orks w hose infrastru ctu res w ould not allow the exchange of similar traffic levels. In essence UUNET believed that shortest exit routing was requiring it to provide natio n al and international data transport, as well as connectivity and support services, to companies that could not provide similar services in return.

6.6 UUNET even went so far as to tell the Federal Networking

Council th at it w ould no longer continue to offer the governm ent "charity" connections and that government networks which had been getting cost free peering would now have to pay for the privilege of their connectivity.11

6.7 UUNET's move was based onrecognizing the true nature of the relationship betw een UUNET and its customers. In one instance, the public shake- up resulted in the dismissal of a netw ork executive, David H olub, th en -p resid en t of Whole Earth Networks Inc. Holub subsequently posted a public message on the North American Network Operators G roup (NANOG) e-mail discussion list arguing that UUNET's refusal to peer was anti-competitive and possibly illegal. H olub argued that Internet interconnection is analogous totelecom m unications carrier interconnection, and that access should be regulated on the grounds that "universal reachability" of networks is in the public interest - and that this regulation should extend to m an d atin g public disclosure and regulatory approval of terms.12

6.8 Holub presented a solution for state public utility comm issions (PUCs) to regulate In te rn et peering arrangem ents, and require disclosure to p rev en t discriminatory practices. PUCs are allow ed to regulate interconnection agreements among interexchange carriers and local exchange carriers under section 251 of the Telecommunications Act 1996 in cases w7hen the market players cannot come to an accord. However, while Holub would like to argue th at the Telecommunications Act applies on the basis th at In tern et interconnection is analogous to telecommunications carrier interconnection, it is generally accepted th at In tern et

backbone providers and ISPs are not classified as common carriers under the Act when they provide Internet services using the packet transmission service of a common carrier affiliate - such offerings constitute "inform ation services" u n d e r the Telecommunications Act.13

6.9 In addition to the disputes betw een Internet backbone providers and ISPs, a further battle has emerged between the backbone providers and major content providers (web farms). Essentially, the web farms have sought to argue that because they provide content which is in high d em an d , they are entitled to interconnection on a peerin g basis w ith the Internet backbone providers.14

7. ALTERNATIVE MODELS TO PEERING - FINANCIAL SETTLEMENTS INTERCONNECTION ARRANGEMENTS

7.1 It is now becoming common practice for Internet backbone providers to im plem ent a financial settlem ent system with ISPs that accounts for the use of each other's facilities for "transiting" traffic. A number of alternative settlem ents m odels are now em erging, along with more traditional t e l e c o m m u n i c a t i o n s interconnection regimes. Such arrangements require that the parties determine a fair and reasonable charge for interconnection - as compared w ith SKA peering which sim ply m andatesinterconnection for free, often w ith o u t recognition of the u n d erly in g value of the network services provided by each party. F urtherm ore, settlem ents arrangem ents provide far greater flexibility, facilitating asym m etrical payment arrangements where appropriate in order to reflect the comparative value offered by each of the interconnecting networks.

10 COMPUTERS & LAW


7.5 There is also considerable diversity in th e various settlem ents arrangem ents implemented, due to added complexities arising from the fact that there is no uniformity as to which way a settlement should go in the case of a traffic imbalance: ie: w h eth er the parties should adopt a "sender pays" or a "receiver pays" approach.

8. RISKS ASSOCIATEDWITH BREAKING AWAY FROM PEERING 84

8.1 Interconnection is absolutelycrucial to the operation of the Internet - more so than for any other network-based industry, given the absence of end-to- end circuits. The delivery of an item of data over the Internet depends on negotiating the successful delivery of each separate packet of data via a complex and p o tentially chaotic web ofinterconnections betw een networks. However, despite the enorm ous technical challenges posed by connectivity on the Internet and the potential for chaos, full global connectivity has now become a reality th ro u g h seamless connectivity among millions of routers, services and users. Furthermore, it has been achieved in a largely unregulated environment.

8.2 Having overcome seemingly insurm ountable technical obstacles to In tern et interconnection, it is the age- old economic and regulatory problems which now pose the greatest threat to the expansionof connectivity on the Internet, 8.5 as regulators are called upon to intervene in the myriad of disputes arising as a result of the break away from SKA peering interconnection arrangements.

7.2 The flexibility inherent in suchfee-based interconnection arrangements enables Internet backbone providers to enter into interconnectionarrangements without losing the ability to be fairly com pensated for their in frastructure costs and ongoing network expansion. Accordingly, suchinterconnection arrangements have the potential to promote a greater level of interconnection in the long term than would otherwise be the case - and to deliver an optim al level of In tern et connectivity.

7.3 In effect, we are m oving tow ards a tiered pricing structure encompassing the full range from retail one-way charges to sym m etrical settlem ent charges to asym m etrical settlem ent arrangements to peering.

Practical implementation7.4 In practice, m any of the

downstream ISPs acquiring connectivity on a settlements basis do not pay based on their actual usage, bit by bit, but based on a usage profile. This is because the costs and practical difficulties associated with metering every data flow simply outweigh the benefits - particularly given concerns that the use of technology req u ired to m easure exact usage could adversely impact on the speed and quality of the In tern et service delivered. However, this usage profile should be distinguished from the alternative "accounting rate" approach which is not cost-based and is inevitably arbitrary, bearing little relationship to traffic levels and the level of netw o rk services provided. The latter approach sim ply serves to aggravate the businessinstability which is already p rev alen t in In tern eti n t e r c o n n e c t i o n arrangements.

Regulatory concerns8.3 The shift away from peering

has, not surprisingly, drawn cries of protest from affected

ISPs which had previously operated very profitable businesses based on the arbitrage opportunities arising from:

• peering w ith In tern et backbone providers in order to obtain Internet access for free; and

• like foreign exchange arbitragers, extracting revenues at the time of on- supplying access to the Internet to smaller ISPs.

However, notwithstanding the protest of ISPs, it is widely acknowledged that:

• i n t e r c o n n e c t i o n arrangem ents have not adequately reflected the actual costs of interconnection in the past;

• the arbitrage opportunities presented by peering resulted in there being little incentive for ISPs to look for efficiencies in the aggregation of traffic or to deliver v alue-added products and services to their customers; and

• while ISPs resen t being delegated to the status of customer, any network will inevitably be a customer of another netw ork w ithin the layered structure of the Internet, and so unable to control its fixed costs, unless it is a very large network with enorm ous capacity and is u p g rad in g continuously (in w hich case it will qualify for peering with another very large network).

Having said this, there are still very real risks that the shift away from peering will attract the attention of regulators, with the following scenarios likely to ring alarm bells:

• The mere fact that Internet backbone providers are in a position to substantially increase the charges imposed for Internet access, with the accom panying

COMPUTERS & LAW 11


pow er to w ithhold connectivity to a portion of the m arket, is of itself sufficient to raise concerns as to the possibility of a degree of market power being vested in Internet backbone providers.

• In the case of a vertically in teg rated backbone provider, the move away from peering can very easily be interpreted as an attempt to put a "financial squeeze" on the Internet backbone p ro v id er's competitors at the ISP layer of the industry.

• Concerns may arise as to possible collusion amongst the "club" of large Internet backbone providers, based on the assum ption th at there is no incentive for them to allow any emerging Internet backbone provider to enter the club. Such concerns do, how ever, indicate a lack of u n d erstan d in g of the commercial im peratives which are associated with peering. Any agreement to peer with an additional network of a comparatively similar size can only add to the value of the existing netw orks w ithin the "club", since it provides those netw orks w ith increased connectivity and access to a greater range of netw ork services - as compared with the desire to keep the sm aller networks out of the club on the basis th at they are perceived to "free ride" w ith o u t p roviding any value to the existing networks within the club.

• Concerns may arise in relation to the imbalance in power between an Internet backbone provider and an ISP, such that the Internet backbone provider is in a position to absolutely refuse to recognize the (albeit lesser) value

provided by the ISP, given that the ISP has a greater need for connectivity to the backbone provider than vice versa.15

• Finally, even if a regulator is prepared to accept that a particular ISP does not bear a complementary cost in terms of network resources, the m ere fact th at the revoking of a peering arrangement could result in the affected ISP being hurt m ore than the* In tern et backbone provider is a disparity which could give rise for concern on the part of regulators.

Possible market pow er vested inInternet backbone providers8.6 Regulators may wish to take a

closer look if an In tern et backbone provider becomes very large, given the concerns th at this could ultim ately result in a greater number of other In tern et backbone providers and ISPs:

• becoming more beholden to that Internet backbone pro v id er to carry their customers' traffic than that Internet backbone provider is by comparison to them; and

• being forced to accept In te rn et services on w hatever term s th at Internet backbone provider chooses to offer, or run the risk that a large proportion of their customers will not have connectivity to the customers of that Internet backbone provider. The result could be a loss of value in the services which they are supplying to custom ers, due to lim itations in the connectivity which they are able to offer.

8.7 The potential outfall of the break away from peering is therefore likely to give rise to concern on the part of the regulators, since the

com petitiveness of thebackbone layer of the Internet is crucial to the competitiveness of the entire Internet industry. A lack of com petitiveness at thebackbone layer would enable Internetbackbone providers to charge above cost, with little incentive to invest in network expansion and upgrades, and would impact adversely on the price and quality of services delivered to end users.

8.8 Concerns have been expressed as to the control vested in large Internet backbone providers - particularly in the U nited States w here the larger Internet backbone providers are going through a period of consolidation, both in terms of business rationalization and by m erger. In tern et backbone traffic is being concentrated into the hands of fewer and fewer organizations and, while the MCI-WorldCom merger focused on the issues of peering arrangements, it did not set any ground rules or guiding principles in this area.

8.9 However, the potential for alack of competitiveness at the backbone layer is in fact far greater in those regions which are still in the start-up stages of infrastructure development. This is due to the propensity for incum bentt e l e c o m m u n i c a t i o n s m onopolies to refuse to recognize the value offered by new e n tran t backbone providers, and to use their m arket pow er to w ithhold interconnection to new entran t backbone providers on fair and reasonable terms.

Australian example; misuse o fmarket pow er by incumbentmonopoly provider8.10 This exact scenario arose in

Australia during 1997/98 with the incum bentt e l e c o m m u n i c a t i o n s m onopoly, TelstraC orporation, refusing to provide interconnection to the

12 COMPUTERS & LAW


new entrant, Cable & Wireless Optus. This was despite the fact that:

• the new e n tran t had constructed a national ATM Internet backbone, with greater capacity and speed th an th at of the incum bent's In tern et backbone;

• the new e n tran t had acquired substantial international capacity in its own right (a key element of any In tern et backbone operation outside of the United States, particularly in Australia where at least 70% of all Internet traffic is international); and

• traffic levels between the netw orks were roughly equivalent.

8.11 The incum bent m onopoly refused to enter into any form of interconnection agreement with the new entrant, insisting instead that the new entrant must become a customer of the incumbent in order to access content and customers via the incum bent's netw ork. This entailed p aym ent to the incumbent of $190 per gigabyte of data received, which was a b u n d led rate calculated to recoup the costs of both national and domestic traffic - despite the fact that the new entrant only sought access to the incum bent's dom estic netw ork. F urtherm ore, the incumbent refused to pay the new entrant on a reciprocal basis for access to content via the new entrant's netw ork, despite the equivalent traffic flows between their networks.

8.12 A stand-off of some 15 months finally resulted in regulatory in terv en tio n in May 1998. However, the outcome of that regulatory intervention clearly illustrates the real dangers of the regulators "g ettin g it w rong". The regulator mandated interconnection not only between the incumbent and the only alternative

backbone p ro v id er in Australia, Cable & Wireless Optus - but also m andated that the incumbent must enter into in terco n n ectio nagreements with two of the larger ISPs in Australia (which happened to be the only other entities w hich had lodged complaints with the regulator in respect of the incumbent's interconnection practices at that time).16 Neither of those ISPs had invested in the construction of th eir own infrastructure, and instead p rovided In te rn et services utilizing the infrastructure supplied by the In tern et backbone providers. However, they managed to convince the regulator th at th ey had equivalent status as Internet backbone providers, largely based on the fact that they had nodes in each of Australia's capital cities.

8.13 The arbitrary nature of the Australian regulator's decision is highlighted by the fact that there were, and still are, other ISPs operating in Australia with larger operations than those ISPs designated by the regulator as being entitled to interconnection at the backbone level. Ironically, one of the ISPs nam ed by the regulator had n o t even acquired its own international capacity - despite the fact that international capacity is a key requirement in the supply of Internet services in Australia, since 70% of all Internet traffic in Australia is international. Instead, the ISP relied on the supply of in tern atio n al capacity by the In tern et backbone providers, thus reinforcing its true status as a custom er of the In tern et backbone providers.

8.14 The Australian regulator has not disclosed the criteria which it applied in reaching this decision, thus adding to the considerable uncertainties which the decision has created in respect of both:

• the relationships betweenIn tern et backboneproviders and those ISPs which were not included in the regulator's order; and

• the ongoing relationships betw een the In tern et backbone providers and those ISPs which were named in the regulator's order, given that the size and scope of operation of those nam ed ISPs is constantly shifting in comparison with that of In te rn et backbone providers and other ISPs operating in Australia - so th at the regulatory intervention runs the risk of entrenching the status of the nam ed ISPs in an en v iro n m en t which is constantly changing and evolving month by month.

9. DIFFICULTIES FACED BY REGULATORS - ASSESSING VALUE OF INTERCONNECTION

9 1 Any regulatory intervention in the area of In tern et interconnection is fraught w ith difficulties, the most obvious being those associated with u n d erstan d in g a fundam entally newtechnology. However, even greater difficulties arise due to the uncertainties associated with attributing value to the various components of Internet interconnection, namely:

• broadband capacity and the other network features and services;

• connectivity (ie: the scope of reachability to end users); and

• other value-added services (including content and other value-added applications).

9.2 A fair price for interconnection is all about determ ining the relative value offered by each network to the other. However, this evaluation involves a

COMPUTERS & LAW 13


number of added complexities in relation to In tern et interconnection.

9.3 R egulators are no longer dealing with relatively similar netw orks which offer com parable capacity and speed, as in the case of PSTN networks. The technology and capacity in co rp o rated in In tern et backbones varies substantially, and there will inevitably be considerable differences in the quality of the network services, features and benefits offered by the respective netw orks particularly in relation to the carriage of m ulti-m edia content.

9.4 Given the failure on the part of Internet backbone providers to take adequate account of the economics of In tern et interconnection in the past, the principles for attributing value to various com ponents of Internet interconnection are still being developed. The criteria applied in the past have been largely subjective (and may continue to be subjective for so long as In tern et interconnection rem ains d ereg u lated ). For example, the notion of "similar size" in a netw ork has been determ ined by traffic flows, num bers of custom ers, backbone capacity, the geographic reach of connectivity, the services and content that can be accessed via a particular network, and other factors which the networks have been prepared to recognize as a basis for the free exchange of traffic:17

• g reater geographic coverage m ight be the determining factor which resulted in one netw ork being prepared to value the benefits derived from interco n n ectio n w ith an o th er as being equivalent;

• a network might perceive that the significant content hosted by another network

had the potential to add value to its own network to th e extent th a t, while infrastructure investment was n o t equivalent, the overall value provided by each network to the other was neverthelessequivalent; and

• similarly, an ISP with a very large client p o p u lation w ith in a lim ited geographic locality might be considered to provide significant value to another network.

9.5 The uncertainties associatedwith the value attributable to each com ponent of in terco n n ectio n areag gravated by the lack of tran sp aren cy in Internet interconnection agreements between networks.18 To date, little has been disclosed or, in fact, u n d ersto o d about the business model assumptions underlying and m otivating i n t e r c o n n e c t i o n arrangements. The only time w hen th ere has been any public discussion as to the dynamics involved and the business case assum ptions u n d erly in g In tern etinterconnection agreements has been when disputes have arisen - so that inevitably there is only lim ited expertise, p articu larly am ongstregulators, in evaluating the stakes involved in such a dispute.

9.6 U nfortunately , there is noindustry consensus on these issues - and nor is there likely to be consensus in the near future. Any view adopted in relation to the values attrib u tab le to the various com ponents ofinterco n n ectio n willinevitably be driven by the economic incentives of the particu lar ind u stryp articip an ts. ISPs w ith o u t investm ent in infrastructure will inevitably seek to bundle the com ponents of interconnection and argue that

each com ponent should be accorded equal value in the interconnection equation. For example, an ISP which is able to offer connectivity to a large custom er base, or access to highly popular content, will seek to argue that this equates to the value offered by another n etw o rk w ith substantial in v estm en t in backbone infrastructure.

Value associated withconnectivity9.7 ISPs argue that connectivity

(w ith o u t routing and transport) should be settlement free. When a customer of one ISP com m unicates with a custom er of another ISP or backbone provider, both custom ers benefit. Each customer pays for use of the n etw o rk to which it is connected. The operators of both networks are paid by their customers, and there should be no further need for them to settle.

9.8 ISPs have sought to link the issue of connectivity with the price payable for access to the In tern et, claiming that the rem oval of peering arrangements serves to impede the overall connectivity on the Internet and is therefore anticompetitive. However:

• while each ISP is able tooffer increasedconnectivity to an Internet backbone (in the form of connectivity to the ISP's own customers), this does not of itself entitle an ISP to peering - there m ust be recognition of the disparate value between the network resources provided by the Internet backbone provider and the increased connectivity offered by the ISP;

• if the lim ited marketing expenditure required to develop a customer base is accepted as being sufficient to provide an entity with free access to a national

14 COMPUTERS & LAW


Internet backbone, this will Vali act as a serious deterrent for g g ongoing investm ent in infrastructure in the long term;

• the refusal to provide peering to certain ISPs does not p rev en t them from obtaining connectivity - it simply requires that they interconnect at a different point in the chain and pay the appropriate price for that connectivity. It is open to an ISP to negotiate such connectivity either:

• on a one-to-one basis - as a custom er of the Internet backbone provider or another ISP, or via a settlem ents i n t e r c o n n e c t i o n arrangem ent which reflects the disparate values betw een the networks; or

9.10• on a many-to-one basis

- by aggregating or peering its netw ork resources with those of a group of similarly sized networks to create a single larger network with increasedbargaining power, which is then in a position to jointly interconnect with the Internet backbone provider or a com paratively larger ISP; and

• the ease w ith whichISPs can aggregate their resources byinterconnecting with other "peers" (whether on a regional or a national basis) is unique to the In te rn et, and results in the technology itself im posing its own constraints on any potential market power of Internet backbone providers - including in relation to their ability to w ithholdconnectivity.

a sso c ia ted w ith con ten tTurning to the value of content, a dispute has arisen in the United States between major content providers (web farms) and major backbone providers in relation to the value attributable to content in interconnection arrangements - as illustrated by the highly publicized dispute betw een Exodus Communications Inc. (a publicly held company with net assets of nearly US$60 m illion) and GTE Internetworking. Exodus is a national web farm hosting such valuable content sites as USA Today and GeoCities, and had entered into a short-term peering agreement with GTE Internetw orking. GTE Internetworking announced on 9 July 1998 that it would not be renew ing th at peering contract.

GTE In ternetw orking's strategy was based on the fact that peering no longer works in the case where an Internet backbone provider is faced with asymmetrical traffic flows from a web farm:

• Peering is based on the assumption of some form of sym m etry-not in the exact number of bytes exchanged in each direction, but by an order of m agnitude. Peering contemplates that both networks will strive to present equal value to each other. Furtherm ore, it is based on the assumption th at both sender and receiver pay an equal share for traffic. The asymmetrical traffic flows destroy the assumption on which the Internet has so far been built, namely that both sender and receiver must contribute to support traffic costs. In this case, Exodus sent many times more bits to GTE Internetworking than were sent the other way, and these asymmetrical traffic flows im posed a heavy

burden on the performance capabilities of GTE Internetworking's network

• The b u rd en on the

backbone provider is aggravated w hen h o t potato routing means that the content is immediately passed onto the backbone p ro v id er's netw ork - resulting essentially in a "receiver pays" pricing model for the Internet. The only solution to this scenario is "best exit routing" (which involves imposing responsibility on the web farm to carry the traffic flow to an exit point closest to the location of the In tern et backbone p ro v id er's custom ers, going some way towards restoring the sym m etry which is destroyed by the web traffic) and settlement based peering.

9.11 In response, Exodus sought toargue that since it was a highly desired content provider, it was entitled tointerconnection on a peering basis w ith large global backbones such as GTE internetw orking - despite a major imbalance in the traffic flows between the networks (estimated to be something in the order of sixteen to one). In effect, this amounted to a claim that Exodus should receive free interconnection because it provided content that GTE Internetw orking custom ers would want to access.

9.12 If Exodus were to succeed in its claim, it would suggest that access to content either has surmounted, or is on the verge of surmounting, the need for equivalence of networks and symmetry in the exchange of backbone traffic. However:

• no consensus has been

reached that web hosting com panies need to be interconnected to the Internet for free - to remain a peer of everyone forever

COMPUTERS &c LAW 15


without being required to pay anyone for an inevitable cost of business, ie: that portion of upstream bandw idth consumed by the data exported;

• such an approach would

lead to a "receiver pays" model, which would have major implications for the In te rn et in d u stry including the inevitable risks of m ulti-m edia spam m ing and the associated costs which that would entail for end users.

Implications fo r the regulators9.13 As noted above, there is no

ind u stry consensus on the issue of attributing value to the various com ponents of interconnection - and nor is there likely to be consensus in the near future, given that the issue will inevitably be driven by the particular economic motives of the various industry participants.

9.14 While decisions relating to interconnection w ithin the commercial environment may in fact be reached on a subjective basis, any regulatory in terv en tio n m ust tread lightly. It m ust also be premised on a sound analysis of the value attributable to the various com ponents of interconnection in any particular dispute. Failure to undertake such analysis will inevitably result in arbitrary decisions on the p art of regulators, as they struggle to steer a p ath betw een the emotive and self-interested claims of the various industry players.

9.15 Internet backbone providersare not yet treated as common carriers, and universal Internet access has not yet become a public policy objective like universal telecommunications service. However, despite the fact th at thetelecommunications industry is subjected to a regulated access regim e (w hile the

In te rn e t is n o t), the telecommunications industry is still subject to the constraints of pricing principles which ensure that the access price is set at somewhere between:

• the total service long-run

incremental cost (TSLRIC) of providing the service; and

• a price which reflects the

opportunity costs (ie: what is forgone by employing resources in thedr current use rather than the most valuable alternative use).

9.16 However, at least in Australia, the recent decision by the regulator suggests that these basic access principles have been thrown out the window in the d eregulateden v iro n m en t of Internet in terco n n ectio n . As a consequence of the decision not only to m andatein terco n n ectio n betw een Internet backbone providers, bu t also to accede to the demands of the two ISPs who knocked on the regulator's door demanding free access to the Internet:

• Internet interconnection in

Australia is now clouded by total uncertainty, due to the distortions created in arbitrarily deeming two of the ISPs (but not others of equivalent or greater size) to be peers of the Internet backbone operators;

• activities at the wholesale layer of the Internet carriage services market in Australia have been severely cu rtailed , w ith a consequent loss of the efficiencies associated with the aggregation of traffic which one w ould normally expect to see at the wholesale layer - so that backbone providers are forced to focus their attention on the potential revenues to be derived at the retail layer of the industry;

• the incentives for

investm ent ininfrastructure in Australia are seriously undermined; and

• In tern et backbone

providers are forced to rely on the bundling of Internet carriage services w ith v alue-added In tern et services and u n related com m ercial services in order to derive a return on their in frastructure investment.

9.17 The conduct of the Australian regulator in deeming two of the larger ISPs to have equivalent status to the Internet backbone providers represents a sledge-hammer approach to regulatory intervention at a time w hen the In tern et industry in this country is still at a very immature stage - so that even "light han d ed " regulation is inherently risky. F urtherm ore, it conflicts starkly with the international trend towards recognizing the u n d erly in g costs of interconnection.

Dilemma between promotion of competition and retaining incentives for investment in infrastructure

9.18 The recent disputes in relation to Internet interconnection bring to the fore a major dilemma for regulators, ie: balancing the promotion of com petition against the m aintenance of long term incentives for investment in in frastru ctu re. Any policy decision in this regard must take account of the particular characteristics of In tern et infrastructure - which are very different to those exhibited in the telecom m unications industry:

• The establishm ent and

incremental costs of ISPs are com paratively m inim al when compared with those of In tern et backbone providers. ISPs m ust purchase a terminal service,

16 COMPUTERS & LAW


a m odem pool, ISDN netw o rk term inating equipment and routers, as well as dial-up lines to the telephone netw ork. Clearly, increases in the volume of customers will result in incremental costs relating to increases in com puter memory, disk space and the number of incoming lines. The ISP's link to the Internet may also need to be upgraded from time to time.

• Some of the larger ISPs may

also choose to acquire leased fibre capacity, in which case they will incur some upfront or fixed costs, with discount incentives enticing them to enter long term contracts. The extent of the sunk or fixed costs d ep en d s on the typical contract length, which is in ^ the range of three to five years.

• Once the ISP has made the

basic investment (and the 10-1 acquisition of leased fibre capacity is not essential), there are numerous means available to an ISP to develop a national or regional network, utilizingnot only the fibre capacity supplied by the Internet backbone provider but also routers and sw itches - which enable the instant establishm ent of virtual private netw orks for a fraction of the cost of constructing the required In tern et backbone infrastructure.

• Accordingly, while the

grow th of the In tern et ultimately depends on the capacity of the fibre in the ground, an ISP m aybe able to exhibit all thecharacteristics of a regional or natio n al netw o rk w ith o u t a com parable investment.

• W hen considering the

application of traditional

telecommunications access principles to the Internet, the ISP's costs are in fact minimal w hen compared w ith those of a t e l e c o m m u n i c a t i o n s reseller which faces the costs associated w ith acquiring access to the local loop before it can commence business.

Accordingly, the discrepancies in the level of capital investment required to operate as an In te rn e t backbone provider, as compared with an ISP, will impact adversely on the ability of the backbone provider to compete against the ISP if the ISP is able to obtain access to the Internet backbone without bearing an appropriate share of the costs associated with that access.

WHERE TO FROM HERE? THE EMERGING SIGNIFICANCE OF CONTENT

It is generally accepted that an ISP w ith large volum es of valuable co ntent is not automatically entitled to SKA peering w ith an In tern et backbone provider, ie: the value of the content is not commensurate with the capital investm ent req u ired to establish backboneinfrastructure19. However, the fact remains that an ISP which provides exclusive access to p o p u lar content is a more attractive proposition to the consum er than one w hich merely provides access to the In tern et generally particularly in a grow ing global m arketplace w here generic, culture-proof popular content increases in popularity to the extent that Internet usage expands across different countries and cultures. F urtherm ore, in the negotiation of an alternative interconnection arrangement involving settlem ent or billing, access to the consumer via popular content is the ISP's

leverage in claiming a right of set-off against the interconnection paym ents owing to a backbone provider.

The rise o f online portals

10.2 The tre n d tow ards recognizing the value of content is m ost clearly evidenced in the rise in po p u larity of ISPs portals, w here ISPs b u n d le their carriage function w ith the provision of popular content. Following the extraordinary success of America O nline, m any ISPs have begun to provide exclusive access to content th ro u g h strategic alliances w ith content providers. Thus in the USA, Disney has invested in Infoseek, and NBC has bought Snap, CNet's online directory service. In Australia, Microsoft has aligned with:

• C hannel 9, Australia's

largest and most popular television broadcaster; and

• Telstra Big Pond, the largest

ISP in Australia,

• to provide a portal to which

Big Pond custom ers can obtain free access.

10.3 The bundling of these brands and the services they offer represents considerable value for the customer. In return, the ISP and its content providers retain a loyal base of customers. Because such customers are less prone to churn, this reduces the required level of ongoing promotional expenditure.

10.4 For the backbone provider, this translates into increased traffic flows of a stable and constant n atu re - to g eth er with consequent increased revenues u n d er the settlem ent arran g em en t. Exclusive content provides even greater benefits to the backbone p rovider in the form of increased traffic flows. If customers choose to use only one portal to access all of their content, the Internet backbone

COMPUTERS & LAW 17


provider will gain - at the expense of those ISPs relying on the lottery of the traditional Internet consumer model.

Using advertising content to fu n daccess to the Internet10.5 Even outside of the realm of

trad itio n al m edia such as Disney and NBC, it is clear that content is becoming a more valuable commodity on the Internet as the industry turns to advertising to fund the delivery of content to the end- user at low or nil subscription rates. Value-added services such as the roam ing email services run by Yahoo, Alta Vista and Hotmail are examples of the utility of content to secure access to the customer at a low-cost to the ISPs - showing that customers will flock to a free service and, once th ere, will p u t up w ith prominent advertising on their screens. This is a competitive advantage over the traditional user-pays m odel, w here a charge for each and every login can represent a significant disincentive to sign up as a subscriber. Furthermore, the real providers of the content - the advertisers - are happy to use ISPs to deliver the content because they realize the significant and still rapidly increasing exposure these services can provide.

10.6 In the long term, it is inevitable that the value associated with advertising will have some bearing on the interconnection rates negotiated as betw een networks, given the ability of advertising to decrease the costs of the end user, while at the same time funding the development of popular sites which attract increased traffic flows.

C ontent assists In te rn et backbone providers to differentiate their carriage services

10.7 Internet content can fluctuate w ildly in term s of its usefulness and popularity, and

hence its value. By com parison, the value of carriage rem ains relatively constant.20 Despite the efforts of Qwest and others to provide differentiation in service levels, carriage rem ains a commodity item with service levels being fairly constant across the industry. On this basis, the ability of an Internet backbone provider to connect with ISPs which host popular content will therefore have the potential to impact directly on the levels of traffic via that netw ork - with consequent implications for the status of th at backbone provider in interconnection negotiations with other networks.

W hy content may not becomeking10.8 As a consequence, one might

expect that the golden rule of the cable industry, ie: "content is king", will be replicated in the In tern et industry. H ow ever there are certain dynamics which have stalled the elevation of content as a driver in interconnection negotiations. With the prevalence of SKA peering throughout the industry until recently, this has served to actually decrease the value of content since:

• increased traffic flows do not result in any increase in the revenues derived by an In tern et backbone provider; and

• in fact, increased traffic

flows sim ply create a burden on the performance levels of the backbone provider's network,

• so that in such circumstances

it is not econom ically viable for the Internet backbone provider to attribute any value to the content held by the ISP. Content which is in high demand impacts negatively on the performance of the netw ork, w ith o u t any

perceived benefits for the Internet backbone provider.

10.9 However, with the trend away from SKA peering and towards s e t t l e m e n t s - b a s e d i n t e r c o n n e c t i o n arrangem ents, we will see op p o rtu n ities for In te rn et backbone providers to derive increased revenues from p opular co n ten t, thus ensuring that the value of content should becom e a relevant factor in the negotiation of any settlements based interconnection arrangements.

11. CONCLUSION

11.1 It remains to be seen whetherIn tern et backbones will be treated as telecommunications in frastru ctu re , w ithaccompanying obligations to provide fair and non- discriminatory terms to ISPs seeking interconnection - or w hether Internet backbone providers will be able to pick and choose with whom they interconnect and on w hat terms.

[Editors' Note: In Australia Internet backbones are considered to be 'telecommunications networks' and Internet access services are considered 'carriage services' u n d er the Telecommunications Act. Carriers and carriage service providers are required to comply w ith any applicable standard access obligations under Part XIC of the Trade Practices Act 1974.)

11.2 How ever, comm ercial imperatives suggests that not all In tern et backbone providers and ISPs will have direct and seamless interconnection with all others

prim arily because commercial interests favour disconnection of ISPs unless and un til they agree to interconnection arrangements involving transfer payments u pstream , reflecting the u n d erly in g costs of the infrastructure provided by the backbone provider.

18 COMPUTERS & LAW


11.3 The recent disputes in relation to Internet interconnection bring to the fore a major dilemma for regulators, ie: the policy decision of w hether com petitiveness should be promoted at the expense of long term incentives for investment in infrastructure. In this regard, facilities-based com petition has been the cornerstone of US, UK and OECD telecom m unications com petition policy and, arguably, is even more im portant in relation to the Internet where we are seeing an exponential growth in the demand for bandwidth. It is only through facilities-based competition that a sustainable and self-sufficient backbone layer will emerge and thrive, with consequent benefits for the entire Internet industry.

[Editors' note: In Australia facilities- based competition has also been a part of the Government's 'full and open competition' policy with any person who holds a carrier licence being able to install telecom m unications

netw ork units for the purpose of supplying carriage services to the public. The requirements for the issue of a carrier licence certainly aren't onerous.]

1 Interconnection on the basis of "Sender Keeps All (SKA) peering" or settlement-free peering.

2 See for example The COOK Report on Internet, May 1998(Vol. 7,N o. 2).

3 Cukier, K .N . Peering and Fearing: ISP Interconnection and Regulatory Issues, presented at Harvard University Kennedy School o f G o v ern m en t's In fo rm atio n Infrastructure Project conference, December 1997. h ttp ://k sg w w w .h a rv a rd .e d u /iip / iicompol/Papers/cukier/html

4 Cukier, K.N. The Global Internet: A Primer, TeleGeography, 1999, p. 113.

5 Ibid, p 115.6 See further section 9 below.7 See further section 6 below.8 See discussion in Cukier, supra at note 3.9 Ibid.10 Economides, N., Lopom o, G., W oroch. G.,

"R egu latory Pricing Rules to N eutralize N etw ork D om in ance", presented at the C o n so rtiu m for R esearch on Telecommunications Policy at Northwestern University, May 1996.

11 The COOK Report on the Internet, Building Internet Infrastructure, January 1998, p. 86.

12 Supra at note 3.13 Telecommunications Act 1996, s 3(20). See

discussion in Frieden, R. Without Public Peer: The Potential Regulatory and Universal Service Consequences of Internet Balkanizion, 3 Va. J.L. & Tech. 8 (Fall 1998). h t tp :/ / vjolt.student.virginia.edu

14 See further section 9 below.15 This is not unlike the situation in respect of

international access to the Internet, where networks outside of the US have been forced to pay the entire cost of the circuits required to acquire international capacity to and from the US. While non-US networks have complained that this practice requires them to effectively subsidize US internet users (to the extent that US users are using those same links to access content outside of the US), the fact remains that those non-US networks need to access content in the US - while US networks do not have the same need to access content outside of the US.

16 The Australian Competition and Consumer Commission issued a "competition notice" pursuant to Part XIB of the Trade Practices A ct 1974.

17 Huston, G. ISP Survival Guide, John Wiley & Sons, Inc. New York, 1999.

18 While there has been much criticism of the " s e c re t d eals" re la tin g to In te rn e t interconnection, as compared with the practice of disclosure in relation to PSTN access arrangem ents, such secrecy is simply driven by co m m ercia l sen sitiv ities. The interconnection arrangements between any two networks will inevitably impact on the margins which those same networks are able to d erive p d rsu an t to their com m ercial arrangements with other networks at other layers of the industry - and disclosure would simply undermine their bargaining power in negotiations with other networks.

19 Refer section 9.20 Supra at note 17, at pp 539-40 where Huston

makes the point that "scant opportunity exists for service-based differentiation".

COMPUTERS & LAW 19

http://ksgwww.harvard.edu/iip/

COMPUTERS & LAW · up a carrier's fixed assets without yielding incremental revenues; and...

Documents

Transcript of COMPUTERS & LAW · up a carrier's fixed assets without yielding incremental revenues; and...