Free Access to Regional Internets: Principal Economic and Technical Issues
Marc LOBELLE <firstname.lastname@example.org>
Xavier BOGAERT <email@example.com>
Paul DEFOURNY <Paul.Defourny@charline.be>
Local information access is what most people mainly need. This holds for all communications media: the telephone system, in which long-distance calls are the minority; newspapers, which produce local editions, etc. When the market penetration of the Internet matches that of the traditional communications media, we can expect a large proportion of its use to be for local information access.
On the other hand, most of the cost of Internet access is currently international, particularly in Europe and in developing countries, where the Internet was originally, and still largely is, mainly an information import market. The structure of the operating cost of Internet service providers shows this domination of long-distance communications (access to [inter]continental backbones).
Internet access to local information is currently as expensive as Internet access to information across the world. The principle of Free Access Regional Internets (Farinets) is to allow free access to regional information, even without Internet subscription, and to charge the user only for access to long-distance information. A Farinet is part of the Internet. Information on the Farinet is accessible from anywhere on the Internet. Users connecting to the Farinet can get information from anywhere on the Internet, but they are not charged if they only navigate within the Farinet.
The paper will explain in more detail the rationale for Farinets. Then, their economic feasibility along with our evaluation of the conditions of their success will be presented. The following section is devoted to the technical solutions used to minimize regional operating costs and to discriminate between regional and global access for charging purposes. The paper will conclude with a presentation of the experience of one Farinet service provider in Belgium and a comparison of approaches.
Information is a consumable product and most of the information a user consumes is regional; even with the possibilities of cable or satellite TV, most people watch local or national channels most of the time. The same is true for radio and newspapers: different editions are produced for different regions. Most of the telephone traffic is also short distance. The only exception to this is the Internet. On the Internet, the location of information sources is not apparent (except perhaps when lines are overloaded!). The Internet was first successful in offering instantaneous access to huge amounts of information all over the world. Internet is a very cheap way to get access so fast to so much information coming from so far. Quite naturally, those who needed this international information were the early adopters of Internet. However, we believe there is no reason to use the Internet only for satisfying this specific need. We believe that offering instantaneous and cheap access to local information would be useful for many more people and would create many new business opportunities.
This project started in the framework of a European program aimed at redeveloping the Hainaut region in Belgium (Objectif 1). A group in the Université Catholique de Louvain proposed to use Internet technology to develop economic activities in Hainaut. This operation is run in the CEDITI (CEntre de DIffusion des Technologies de l'Information), a technology transfer center in information technologies set up in Charleroi by the three universities of Louvain-la-Neuve (UCL), Brussels (ULB), and Namur (FUNDP), with support from Objectif 1. Both intranet-based applications and the development of Internet access and services are involved. This paper deals with the second field.
To develop economic activities based on the use of Internet technology for regional purposes, two main obstacles had to be overcome:
In order to develop economic activities in the region around Internet access, it was thus necessary to make it possible both for end-users to access the Internet at the lowest possible fee and for local companies to gain a competitive advantage by using the Internet (more business opportunities by advertising products on the Internet, more efficient operation by using the Internet for information transfer, etc.) and by doing so locally (i.e., not by advertising one's products on a server located in California!).
Our problem was thus threefold:
Besides, we wanted our initiative to be viable in the long run. We could afford the cost of specific technological developments in the frame of the Objectif 1 project, but not to set up something that was inherently money-losing.
In order to make the Internet attractive to information users, and since most of the cost of Internet access is in international lines (in particular, transatlantic ones in the case of Belgium), we decided to evaluate the feasibility of offering local access for free and global access at the lowest sustainable cost. This bundle of Internet services required the deployment of what we call a Free Access Regional InterNET or Farinet.
Offering local access for free means that the user has to pay the local telephone calls to a local access point but no extra fee to the Farinet service provider (FSP). This gives him free access to all the "regional" information made available on the Farinet (we will call them "Farinet sites"). If the user is interested only in this local information, he does not even have to be registered at the FSP. The idea is to let the costs of this local service be supported by the information providers targeting this regional "audience."
For offering global access at the lowest cost for the user, after testing a few other schemes, we settled on a formula we called "Don't use, don't pay." Registered users of this service are charged only for the days when global access has actually been used, i.e., when through the Farinet they had access to Internet sites outside the Farinet. In this scheme, these days are charged at the daily equivalent of the monthly rate of competing providers. Neither days without Internet access nor days when only Farinet sites (including newspapers, local collectivities, local companies) are accessed are charged.
In order to attract information providers to the project, they are offered Internet hosting at rates similar to those of competing providers but with the bonus of being accessible for free in the region, even by people without Internet subscription.
This means mainly that WWW servers set up on the Farinet are visible both to the nonpaying regional users and to anybody on Internet.
Setting up a self-supporting free service requires a careful analysis of costs, methods to keep them down without hampering the quality of service, and finding sources of revenue higher than these costs.
A Farinet service provider is a special kind of Internet service provider. The costs of Internet service providers fall mainly in five categories: manpower, backbone access, ISP's own communication infrastructure, servers, and user access infrastructure.
Traditionally, backbone access is the largest cost item of an ISP in Belgium. Except for large providers who can afford unidirectional satellite channels, backbone access is bought with the same bandwidth in each direction. The use of bandwidth is thus optimal when inbound and outbound traffic volumes are similar. Since information consumers (end users) use more inbound bandwidth and information providers (servers) use more outbound traffic, it is important to balance these two kinds of "users." Offering free regional access is not related to this backbone access cost. This rule of balancing consumers of global information with information providers in order to even inbound and outbound traffics applies thus equally to traditional ISPs and providers of Free Access Regional Internets.
The communication infrastructure of a Farinet service provider serves two purposes: giving access to the regional servers on the Farinet and giving access to the global Internet. The cost structure of a traditional ISP is based mainly on this second need and the costs of the ISP's own communication infrastructure are included in the cost of access to the global Internet. If access to regional servers is to be offered free of charge, it should induce as little extra cost as possible. One helpful element is that the cost per kilobyte of regional bandwidth goes down quickly with the amount of bandwidth you buy, while the cost of backbone access grows more linearly with bandwidth (unless you really use a lot of backbone bandwidth, then you don't buy any more backbone access). Extra regional bandwidth is thus not very expensive. Needs for extra bandwidth for access to regional servers can be further reduced by caching the contents of these regional servers in local proxies in the access points. However, the use of these proxies must be enforced transparently without any user intervention: experience shows that nothing can force users to configure the use of proxies in their browser. The techniques used to transparently force access to regional servers through local proxies will be discussed in the section on technical issues. Such transparent proxies can definitely reduce the extra cost of the FSP's own communication infrastructure induced by the free service. Besides, these proxies improve response times.
Manpower needs for network operation and server costs do not increase because of the free regional access. They will thus not be discussed here.
Access infrastructure costs will definitely increase with the free regional access since more modems or other access devices will be needed. Fortunately, there is no more risk that people will remain permanently connected (and so block modems and deny service to other users) than with traditional unlimited "flat fee" subscriptions. In both cases the obligation to pay the local telephone communication is the regulating factor.
The actual cost of the free regional access is thus in the access infrastructure, the local transparent caching proxies, and the extra regional bandwidth, if any. The balance between the last two items must be optimized for the lowest cost and best performance.
If the information consumer does not pay for the extra cost of free access to regional information, then the information providers must pay.
Three categories of information providers can gain advantage from the free regional access:
We identified five methods to make Farinets immediately attractive to information providers.
The specific technical problems of Farinets, as presented above, are
WWW, FTP, DNS, and MAIL servers are unmodified ones and are located in the network control center of the Farinet. E-mail users are charged only a small flat fee for the setting-up and the maintenance of their mailbox. They need thus to be registered only on the MAIL server. Traffic from this machine, like that of any Farinet server, is routed anywhere on the Internet. No other specific handling of e-mail users is needed.
All the other technical problems are solved in the access servers.
Because of the nonstandard services to be performed in the access servers, these could not be off-the-shelf devices like those used by most ISPs. The access servers had to be reliable and easily configurable systems and they had to be based on powerful but cheap computers. We decided to use high-end PCs with the SOLARIS X86 operating system. SOLARIS  is very well suited to remote administration. Each machine has a multi-serial line interface to a rack of modems and provides the transparent proxy service for the modems connected to it. This is a very scalable solution: if the load of handling the modem lines and the proxy rises beyond the possibility of the computer, a second one or more can be added.
All the information offered free of charge is duplicated on each access server. This information comes from the central servers of the Farinet or from outside servers paying to be freely visible on the Farinet. Because this information is held in proxies, it is refreshed automatically. Each time a user requests a page, its validity is checked, but in most instances the page will be already be available locally: the free services do not contribute much to the traffic on the communication lines between the access server and the network control center. Since most information is available locally, response times are only limited by the round-trip delay for checking the validity of the page and by the modem bandwidth (assuming the access server itself is properly dimensioned).
User discrimination and transparent enforcing of going through the proxy are performed in the modem line interface software as explained below. The PPP protocol is used on the modem line.
When a user tries to connect, he is first authenticated. The principle is that registered users must authenticate themselves by their name and a valid proof of identity while unregistered users may give any other name and any value as proof.
As for most ISPs, three authentication schemes are accepted: unix login, pap, and chap. However, in the Farinet, if the user name is not in the list of registered users, it is replaced, along with the proof of identity (password or response to the chap challenge) by the name and proof of "the anonymous user." This manipulation is performed on incoming traffic from each modem in a filter located before the PPP module. Beyond this filter, traffic enters a classical PPP package which authenticates the user. If a registered user has given a faulty proof, he is rejected. Unregistered users are always correctly authenticated as "the anonymous user."
Before using this scheme, unregistered users and registered users with a faulty proof were treated alike and were considered as "the anonymous user." The default of this earlier scheme was that the symptom of a faulty authentication by a registered user was denial of access to the global Internet, which looked like a router failure in the Farinet instead of an authentication failure.
The PPP package includes a filter denying to "the anonymous user" access to IP destinations that are not to be freely visible on the Farinet. This filter can be considered as a reversed firewall protecting the Internet against Farinet users. Actually, because of the proxy, the "anonymous user" never needs to get out of the access server except to connect to the mail server, if they have a mailbox (e-mail identifications are distinct from PPP identifications).
When getting out of the PPP module, traffic goes through a second filter, which replaces IP destination addresses by that of the access server if the destination address corresponds to a server cached in the proxy. This redirects transparently all accesses to the sites that must be freely visible towards the local caching proxy Web server in the access server itself.
The PPP package keeps detailed logs, including identification of the sources and destinations of outgoing IP datagrams and correspondence between account names and source addresses. Offline processes analyze these logs to provide accounting information.
The discrimination between Farinet sites and external sites has one little inconvenience because Farinet WWW sites may (and often do) include URLs to external sites, most of them not freely visible. Because of this, users visiting the Farinet WWW sites may try to go global without knowing. If they are unregistered "anonymous" users, they will get a not-so-friendly error message telling them that the requested server is not available. If they are registered users of the "Don't use don't pay" service, they will enter a paying day even if they did not intend to. It would be better to display a gentle warning message each time a nonpaying user is denied getting out of the Farinet and another (optional, with a button to ask confirmation) for the registered users when they are entering global mode.
These gentle warning messages must be issued automatically by the proxies since it may not be expected that all authors of WWW sites visible on the Farinet will add something like a filtering cgi on each outgoing link.
The technology needed to implement the Farinet has been developed by the Université Catholique de Louvain and the CEDITI.
A company called Charline Productions s.a. was created to operate a Farinet in French-speaking Belgium. It is owned by a consortium of press groups totaling nearly 10 newspapers, progressively available on Internet, localized or visible on Charline (one of them, Le Soir, is reputed to be one of the most visited French-speaking sites in the world).
The technical operation of the network is subcontracted to the CEDITI.
The infrastructure is currently based on three access servers. Given the charging policy for telephone calls of the Belgian telecom operator, Belgacom, these three access servers allow access to the Farinet at local telephonic rates from nearly all French-speaking Belgium (about 4 million inhabitants). The central site of the Farinet, located in Charleroi, includes the information servers, the interface to the global Internet, and the network control center. All access, mail, and information servers are PENTIUM pros with SCSI peripherals.
The fact that services are offered locally and better-than-average technical expertise is available to Charline through its links to the CEDITI proved to be really competitive advantages.
Charline has now been operating for about two years, first only in Charleroi, then progressively in most of French-speaking Belgium. At the time of writing, the company is not yet making enough money to cover all its costs but according to the trends, the company should do so during 1998. However, the losses have always been very limited and would probably have been higher if Charline had been operated as a regular ISP instead of as a Farinet service provider. Indeed, operating costs would have been the same and some clients of the hosting services would probably have selected a "bigger" provider if Charline hadn't had the advantage of the free regional access.
The idea of free Internet access has been applied in several places, usually subsidized by local authorities, as in northern France, but so far we have found no other case of a commercial provider offering free regional access. The reason is probably that the success of this approach requires, as explained above, the fulfillment of several economic conditions and technical developments that may not be underestimated.
Free Access Regional Internets are a new possible use of Internet technology. They are technically feasible and economically sound. They can stimulate local markets and some day, perhaps, reduce the amount of junk mail distributed to our houses' mailboxes when Farinets will be widely used to distribute local information.
This project has been supported in part by the European Union and the Walloon Region in Belgium in the framework of the Objectif 1 programme for Hainaut.
Beside the authors, many people have contributed to this project: in particular (in alphabetical order), André April, Guy Detroz, Paul Hespel, Pierre-Alexandre Losson, Philippe Nothomb, Louis Piret, Claude Samain, Marie-Claire Schayes, and Bruno Springuel.