n-1-3-040.15 Dealing with Internet Growth - Internet Standards Activity by Craig Partridge There are a number of technical trends coming together right now, all of which turn out to affect primarily IP (the internetwork layer protocol): (1) the explosive growth of the Internet is depleting the IP address space -- so the address space in headers has to grow; (2) the explosive growth of the Internet is simultaneously stressing the size of on routing tables for IP routers, and the general routing architecture -- so the IP architecture for routing and addressing needs to be revised to reduce routing stress; (3) DARPA-sponsored work on multimedia conferencing via IP is coming of age (witness the delivery of the IETF plenaries to over 200 sites worldwide via the Internet last month) -- this work is believed to eventually require some modification to IP to support the provision of some form of performance guarantees (for example, some of the sites on the IETF multicast experienced delays of several seconds -- too long for comfortable conversation). At the same time, there is no firm agreement on exactly when each of these trends will become critical to solve. The IP class B address space is due to run out (I believe) next year, but there is an interim plan called CIDR which is supposed to get us past that problem (CIDR is completely transparent to hosts). Exactly when CIDR will falter is unclear, but folks assume in a few years. The routing table explosion seems more severe but again, timetables for disaster vary. A couple of years, more or less, is a reasonable guess. The multimedia work is still embryonic. Complete proposals for architectural extensions to IP are not expected until early next year. Agreement on what extensions work best is probably a couple years in the future. Now, the key problem is timetable. We need to develop addressing, routing and multimedia extensions good for an Internet of 100s of millions of hosts (perhaps billions). All three extensions are *hard* problems. The IETF has been looking at addressing and routing for over two years (since some folks started graphing the address consumption curve and figured out doomsday was impending). Folks have been looking at the multimedia problem for nearly a decade. And these extensions must be tested thoroughly before launching them on the TCP/IP market. The result is the following muddle. To do serious testing and development of routing and addressing proposals, one needs to decide among the various proposal soon, in case worst-case predictions about addressing and routing stress come true. So the IETF is pushing to try to make decisions by late this year. Part of the motivation behind the IAB's Kobe announcement was a concern that the IETF was not moving fast enough. At the same time, if the IETF decides too soon, it may have to revisit its decision when multimedia is ready to be added to the mix. But not enough is known about the exact form of the multimedia solution to factor it in now. (Past experiences with guessing what is required have shown guessing is a bad idea). There are currently four proposals before the IETF: TUBA -- a proposal to replace IP with CLNP (ISO connection-less internetwork layer protocol). CLNP has bigger address fields than IP but is otherwise architecturally similar to IP. The most common gripe about TUBA is that it probably causes the greatest amount of code to change (because the packet format is completely different, even if the functionality is the same) and that while the addresses get bigger, the larger addressing and routing problems aren't addressed. The TUBA proposal can be found in RFC 1347. Nimrod -- a more radical proposal by Noel Chiappa. Yet to be fully committed to paper but to be taken seriously as Noel is an acknowledged wizard in routing and addressing. Noel believes that he can avoid changes to host software yet provide the extensibility in routing necessary to support multimedia. It is hard to assess the proposal fully. PIP -- a proposal by Paul Tsuchiya, another acknowledged wizard. It is available as two Internet drafts (draft-tsuchiya-pip-00.txt, draft-tsuchiya-pip-overview-01.txt) IPAE -- a proposal to encapsulate larger IP addresses within a sub-layer above IP. Has the merit that hosts can function unchanged in most situations, and may require the least modification to existing code. Major concern is that non-upgraded hosts will exist in disjoint islands of IP, which cannot communicate with each other. Available as an Internet draft (draft-crocker-ip-encaps-00.txt) It is expected the IETF will be deciding among these proposals at the end of the year.