What comes after TCP/IP?

The Internet Engineering Task Force maintains the Internet's protocol standards, including TCP and IP, or TCP/IP, which were first conceived 27 years ago. So isn't it time we asked IETF what's after TCP/IP?

The Internet Engineering Task Force (IETF) meets thrice yearly, and recently for the 48th time in Pittsburgh.

IETF maintains the Internet's protocol standards, including TCP (Transmission Control Protocol) and IP (Internet Protocol), or TCP/IP, which were first conceived 27 years ago during Vint Cerf's summer seminar at Stanford.

So isn't it time we asked IETF what's after TCP/IP?

Today, TCP/IP is IETF's hammer, and every network looks like a nail. This is understandable given the Internet's success, thanks to those who attend the IETF's meetings. See www.ietf.org.

Eight years ago, IETF gathered seven proposals -- not for what's after TCP/IP but for IP's next generation (IPng). In 1994, an IPng was chosen and named IPv6.

The version of IP then in use, IPv4, was thought to be running out of its 32-bit addresses. So IETF planned a smooth migration to IPv6, with its 128-bit addresses, sometime between 1997 and 2001.

IETF worried in 1994 that three new markets -- nomadic computing, entertainment, and device control -- were threatening to diverge from TCP/IP. IPv6 was to provide addresses aplenty; more than a thousand for every square metre on Earth.

The transition to IPv6 is going slowly -- IPv4 still rules. Cisco Systems, Microsoft, and the major Internet backbone providers have not moved over to IPv6.

But cellular telephone companies are talking about putting their next billion mobile phones -- over the next 36 months -- on an IPv6 backbone, even if they have to provide it themselves. But IPv4 vs. IPv6 is an argument over small details. Isn't it time for something entirely different?

Instead of HTML over HTTP over TCP over IP over ATM over SONet (Synchronous Optical Network), how about some real protocol disintermediation -- like Napster directly on the lambdas of DWDM (dense wavelength division multiplexing) fibres?

Just kidding. Evolutions in, under, over, around, and through TCP/IP are likely to proceed in parallel.

Evolution in applications will continue. The Web transformed the Internet, and now Napster is doing it again, without changing TCP/IP, so far.

Evolution of Web plumbing will continue. We are already moving from HTML to XML, without changing TCP/IP.

Evolution of Internet plumbing will continue through new implementations of old protocols. For example, although today's routers ignore IP source addresses, they will eventually authenticate them, without changing TCP/IP.

Evolutions above IP will continue, say with new "in-band" protocols to carry streaming and/or broadcast and/or mobile and/or device data without TCP. Say, SCTP (Stream Control Transmission Protocol) over IP is well along at IETF.

Evolutions will continue in "out-of-band" protocols, using IP over point-to-point media to set up the carriage of streamed broadcast content on IP-free broadcast media.

Evolutions in IP will continue. So let's get to IPv6 already. See www.ipv6forum.com and www.6bone.net.

Consider that packet switching began to make sense three decades ago, as computing became cheaper than communication.

That was about the time computing became cheap enough to squander on interactive time-sharing. But now communication is getting cheaper faster than computing. As time-sharing gave way to personal computing, maybe packetising flows will give way to carrying them continuously.

Maybe TCP/IP, like the PC, is passe. Maybe we actually will go to Napster over lambdas -- just kidding, again.

Or maybe there will be evolution below TCP/IP. Perhaps DWDM's lambda switching will take hold -- with TCP/IP doing very little on top -- the way today's switched Ethernets can, but don't retransmit packet collisions.

Or maybe the Internet will increasingly interconnect new local non-TCP/IP networks. Right after that, the IP-free networks get hooked up using non-TCP/IP backbones, thus ending TCP/IP as we know it. Maybe.

Join the newsletter!

Error: Please check your email address.
Show Comments
[]