38330 Montbonnot-St-Martin, France
Extended abstract for the 5th CaberNet Radicals Workshop,
5-8th July 1998, Valadares, Portugal.
The current set of protocols used on the Internet, IPv4 (Internet Protocol version 4), is on its final legs. It is exhausted, timeworn, and more importantly, it will soon run out of its 232 IP addresses1. The number of hosts that have an allocated address on the Internet is doubling every year. Although proposals such as Classless Inter-Domain Routing [CIDR] are expected to extend the life expectancy of the IPv4 address space for some years to come, a new solution is required. Today, researchers are designing the next generation of Internet protocols: Internet Protocol version 6 (IPv6) [IPv6].
The Internet is to be updated from IPv4 to IPv6 step by step, one host at a time. The changeover is likely to take a decade, and it is possible that some hosts may never switch to IPv6. Fortunately, the specification of IPv6 provides for backward compatibility with IPv4, as well as new services such as real-time flows, mobility and security support, and has provisions for the inclusion of future enhancements and modifications. IPv6 has been designed to last, and the address space has been expanded to identify, in theory, up to 2128 interfaces; this works out to be 340,282,366,920,938,463,463,374,607,431,768,211,456 addresses, which is approximately 665,570,793,348,866,943,898,599 addresses per square meter of the surface of the Earth [IPng]; more than enough to identify every object you could ever dream of!
In this paper, we review some of the future applications proposed for the Internet, and we show that their mobility requirements are not handled well by the IPv6 address space organization. We then propose a new address space organization that facilitates easy development of new mobile applications under IPv6 by directly reflecting the current geographic location of devices 2 in addresses. Our solution requires a constellation of satellites that track, locate, and allow prediction of future locations of mobile devices.
The fifty past years have seen an exponential growth in the sciences. This has lead to tremendous changes in our society, and this trend is likely to continue. We are at the beginning of an extraordinary, revolutionary step in telecommunications in which the Internet will play an important role provided it is capable of accommodating new techniques as well as all types of communication media (e.g., voice, image, text, real-time, interactive, etc.). The Internet should also be prepared for an exponential growth in requested bandwidth, and the number of connected devices, including a large number of mobile hosts.
Today, voice is mainly carried by phone lines on wired and wireless links, television broadcasts are carried by satellites, and the balance of electronic data is carried by the Internet. However, the distinction between TV, phone and text is diminishing: there are many current projects that plan to replace phones with video-phones, TV with interactive user input, and to make 3D-images and sound standard in Web-based communication. The distinction between these services, therefore, no longer depends on the physical medium being used, but rather from the quality of service required: e.g., real-time, interactive, or bulk transfer. For the user, all physical media will be integrated into one logical network, namely the Internet, or whatever network is in place that can meet the requirements of future telecommunications applications.
The integration of communication media will give birth to new tools. Many of today's objects will incorporate an Internet interface, and we will not be limited by the number of addresses as we have already seen in the previous section. Among possible uses, a watch could automatically receive an Internet "time packet" that gives accurate time for the current time zone; cars will be on-line to discover directions to the nearest petrol station; and so on. The trend towards the digitization of data allows for all data types to be carried over a single medium. People will no longer need to buy a TV, a stereo, a tape recorder, a CD player, a fax, etc; a single device connected to a network suffices, and may be more widespread than phones or TVs. Because everything will be accessible through the Internet, people won't need to carry with them photos of friends and music they like. When away from home, anyone could listen to the music of their choice, display a remotely-stored image locally on screen-walls, and supervise the sauerkraut nicely simmering in the kitchen. People will no longer be limited in their movements by the location of their stored data and mobility will then become the norm.
There is a need for data transfer to remain transparent to the user, who should be able to move freely, communicate, and without having to carry a bulky computer. Instead, mobile computers could be incorporated into light helmets, with video showing directly into the eyes, and a microphone and headphones for audio. It is even realistic, to some extent, that a neural-interfaces will be implemented in the human body in charge of translating thought into IP packets. The dream of telepathy would come true, with the Internet as the medium conveying digital thought. Politicians would broadcast their speech to everyone on Earth (god forbid ); Grandma would multicast her wishes to everyone in the family; Romeo could send a valentine, a voice message, or the image of a bunch of flowers, and may be even the smell of roses directly to his love, and only her (provided enough security is added to the Internet so that nobody is aware that he is having an affair); and someone in danger could send an anycast "call for help" message.
Whatever services are provided by the new applications to come, one fact is certain: as the number of applications and users on the Internet increases, the bandwidth that is desired will also increase. The question is, will the Internet be able to offer enough bandwidth to carry all of this new data? New physical medium will have to be used to increase the available bandwidth. Radio waves and infra red will one day be saturated, and it may not be feasible to always bury new optic fibers everywhere. Perhaps, IP packets will then be broadcasted into the water of rivers or ocean, or embedded into electric wires, and applications will be able to choose the best medium according to the type of traffic and the location of the destinations.
In tens of years, many people will live and work on the Moon and in some orbital stations around the Earth. Maybe some on Mars, and in some spaceships anywhere in the solar system. All will be connected to the universal Internet. Obviously, communication will be delayed because of the distance, and interactive communication will not work, but video, the web, etc., should work properly (using something other than TCP, which won't work because of time-outs and the window size [TCPSAT]). The point is, not only nodes will be mobile, but also networks, sites, domains, everything.
Every IPv6 device should be seen as mobile: sites (the orbital station), subnets (the LAN in the High Speed Train), the router (embedded in the satellite), and devices (the IP-watch at the wrist). Although the ability to handle mobility well is a basic requirement in the IPv6 specification, it will not be able to meet those needs because the current solutions for supporting mobility do not scale well, and consider mobility as an exception.
Address Space Organization: The current address space organization in IPv6 considers mobile nodes are the exception, not the rule; it is better to break down this theme, and consider non-mobile nodes as mobiles that do not move, or at least make a strict distinction between the two. Similarly, current work on mobility has only addressed the question of mobile nodes, not mobile networks. There is only some little work that considers mobile networks like embarked networks in a train, a boat, but that is all.
We therefore propose to design a new address space organization which actually reflects the current geographic location of devices, and does not need home agents. The 128-bit address space is cut in two: first, the geographical location part (GPA), which is unique to the hosts location on the Earth, and second, the global identification part (IPA), which is globally unique in the Internet. Our proposal is based on existing systems: the GPS (Global Positioning System) [GPS] which provides the actual location of devices4, and the DNS (Domain Name Server) [DNS1], which provides for name-to-address mappings 3.
We propose a constellation of LEO (Low Earth Orbit) and GEO (geostationary) satellites moving at various heights towards various directions at various speeds. Each LEO satellite represents a mobile domain. With the help of GPS, it is possible to foresee the next locations of the mobile device based on its speed and acceleration. The mobile finds the LEO satellite with the closest trajectory so that it could remain under its coverage a longer time. Every time the mobile switch to another LEO, (or GEO if it decides to remain static), it registers with the Name Server of the Satellite and updates its name-to-address mapping at his home domain (the one where it got its name from) with a record specifying the domain name where it is located. Cache timeouts of DNS records have to be computed according to the frequency of the move.
Correspondents willing to communicate with mobile devices find their location by querying DNS. From the record in the home domain, DNS finds the mobile's current domain name. The reply contains the IPA of the mobile, the domain name, the domain name's position at some time, and details about its speed and trajectory. Since LEO satellites are moving at constant speeds, it is easy to foresee their current location at anytime. A source sends packets to the geographic location where the mobile domain is likely to be next, taking into account propagation delay. The destination address is filled with the IPA of the destination device and the GPA that represents the geographic location of a base station, a router, a subnet or the domain depending on the moving speed of the destination. The IPA of a device is constant, unique, and is obtained from the Internet domain in which the device first registered its name.
Once the location is obtained, routing will be optimum. We therefore see that there is no need for a Home Agent. Of course, using a Home Agent is still permitted for those mobile devices wanting to keep their location secret. In [NAV97], the author proposes a routing algorithm that allows a router to determine if a packet is destined to a geographic location under its geographic area of coverage, or if it should be forwarded. With this proposed scheme, it is easy to send messages to all the people currently residing under a specific geographic area. This is not possible in today's Internet since all people residing in an area subscribe to different service providers.
In summary, GPS and DNS are used to locate Internet devices; devices are identified by their IPA, while the GPA is used for routing directly to the actual or foreseen location of the device. Our address space organization does not have to replace the current one---both could coexist by setting a particular prefix in the packet's header. Our proposal allow mobility of networks and domain in a very transparent fashion. Mobility is naturally supported and there is no need to organize the address space: it is naturally geographically organized.
The reader might think that this would take tens of years before such systems and devices are available and that a new upgrade of IPv6 will already have been designed then. Nonetheless, IPv6 is targeted to play a big role during the next tens of years and is designed to last and being able to give more functionalities gracefully if desired. IPv6 is therefore the candidate internet that will have to accommodate to such new communication needs. We are not going to design a new Internet in twenty years because IPv6 does not provide enough facilities. The specification is not frozen yet, and there is still some debate about the address space organization. It is therefore still time to give full support for mobility.
1 The address space organization consumes addresses; much less than 232 are actually available.
2 We make a distinction between a node, a device, and an interface. A device is a subset of node, it is an entity which may be removed from the node and still exist. It likely has several interfaces; each interfaces is identified by an IPv6 address.
3 DNS is a distributed and hierarchical database in charge of retrieving the IP address of a machine given its name. There is a Primary Name Server and at least one Secondary Name Server (i.e. a duplicate) in each Internet Domain. Name Servers maintain local name-to-address mappings, and likely cache data from queries to Name Servers of others domain. For example, if a user in domain tees.ac.uk wants the numeric address of guadalupe in domain inrialpes.fr, the Name Servers from various domains are called recursively until Name Server in domain inrialpes.fr returns guadalupe's IP address (18.104.22.168).
4 GPS is a positioning, surveying and timing tool. It is made of 24 geostationary satellites.
[AWE91] Awerbuch, Baruch and Peleg, David, "Concurrent Online Tracking of Mobile Users", Proc. ACM SIGCOMM, 1991.
[CAS98] Claude Castelluccia , "Toward a Hierarchical Mobile IPv6 Architecture", eighth IFIP Conference on High Performance Networking (HPN'98), Vienna, September 1998
[CIDR] V. Fuller, T. Li, J. Yu, K. Varadhan "Classless Inter-Domain Routing (CIDR)", RFC 1338, September 1993.
[DNS1] P. Mockapetris, "Domain Names - Concepts and Facilities", RFC 1034.
[DNS2] P. Vixie, S. Thomson, Y. Rekhter, J. Bound, "Dynamic Updates in the DNS", RFC 2136.
[GPS] Global Positioning System, http://ares.redsword.com/GPS/.
[IPng] Robert M. Hinden, "IP Next Generation Overview", http://playground.sun.com/pub/ipng/html/INET-IPng-Paper.html.
[IPv6] S. Deering, R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", RFC 1883, December 1995.
[NAV97] Navas, Julio, C. and Imielinski, Tomasz, "Geographic Addressing and Routing", Proc. of the Third ACM/IEEE International Conference on Mobile Computing and Networking (MobiCom'97)".
[TCPSAT] TCP Over Satellite (tcpsat)
working group at IETF, http://www.ietf.org/html.charters/tcpsat-charter.html.
5th CaberNet Radicals Workshop