Intelligent or stupid? Dumb question

Awhile back I was asked to comment on the”‘intelligent network vs. stupid network” debate. Not familiar with it? It all started in 1997 with the publication of a paper by David Isenberg, then an AT&T researcher, called Rise of the Stupid Network. Isenberg was protesting a recent initiative by his employer to market “The Intelligent Network.”

Isenberg wrote, “The Intelligent Network was a telephone company attempt to engineer vendor independence, more automatic operation and some ‘intelligent’ new services into existing network architecture. However, even as it rolls out and matures, the Intelligent Network is being superseded by a Stupid Network [the Internet].” It’s a fascinating and well-written paper.

Rather predictably, this did not go over well with Isenberg’s employer, and the two subsequently parted ways. But the idea caught fire. It became an article of faith that the Internet is a “stupid network,” characterized by stupid switches, intelligent end points and plentiful bandwidth – and that “stupid” (the Internet) was better than “intelligent” (the public switched telephone network).

Seems like we’re back to the future all over again. Another large network company, in this case Cisco, has begun promoting the concept of “intelligence in the network.” (A March 6 news briefing purported to “examine the trend for more . . . intelligence to be delivered over the network.”) What’s going on? Have the folks at Cisco lost their minds? Don’t they know that stupid is smart and intelligent is dumb?

Oh, for Pete’s sake. With all due respect to Isenberg, the debate has nothing to do with intelligent vs. stupid, and everything to do with carrier hubris and inability to recognize that the assumptions on which it had created a highly profitable business (scarce bandwidth, primarily voice traffic) no longer applied.

Every network has design assumptions built in. You can’t escape it. For example, the current Internet assumes that “end stations” are machines and not users; that data is sent in the form of sessionless packets; and that obtaining the correct IP address for those packets is a function that occurs in a “higher layer” (specifically via DNS, which is very far from a “stupid” function, but indisputably part of the network).

But there are signs these assumptions are a tad frayed at the edges. In today’s world of multiple access technologies and smart phones, a user might connect to the ‘Net at multiple points simultaneously (from a cell phone and a laptop, say), which can do funky things to routing protocols. We’ve all noticed the uptake in interest in services such as VoIP, which require sessions.

The bottom line is that the Internet already contains plenty of intelligence, such as routing, addressing and session-creating. It will need to contain more to handle next-generation services. And that’s fine. The issue isn’t about “intelligent” vs. “stupid” – it’s about how well a set of design principles can map to a changing set of expectations and requirements.