Aug 19

Tim O'Reilly

Tim O'Reilly

Round 2: The Internet As Network of Networks

The other day, I was explaining to a reporter how I could be lumping in cellphones and the next generation of sensor networks into Web 2.0. "Well, Web 2.0 is really not just about the web. It's really about the next generation of internet applications, and includes things like P2P file sharing and VoIP, which aren't based on the web at all. And actually, now that I mention it, it's really not even about the internet, narrowly defined as a class of TCP/IP-based networks. It's really about the internet as it was originally conceived, as a 'network of networks.'"

Those of you who were around in the 1980s will know just what I'm talking about, because that was originally what people meant by "the internet." The term literally came into wide use to refer to a whole set of distinct but increasingly interoperable networks: the ARPAnet, CSnet, DECnet, the UUCPnet, EUnet, NLnet, FIDOnet.... In the late 80s, we actually published a book called !%@: A Directory of Electronic Mail Addressing and Networks, which covered how to address email across 190 distinct networks. (The title !%@ was homage to some of the many special addressing characters that were used before the @ crowded out all the others.) The inter-net was the interoperable network that came to connect them.

It's good to remember this broad definition of "the internet," because the internet is not just about TCP/IP, though it is about the principles that made the TCP/IP based network win out over all the others, and become the lingua franca of interoperability that it is today. We're pushing the boundaries of the old internet, as it comes to include the cellphone network, telematics networks, and other emerging forms of connectivity.

So let's ask, where else can we apply the principles that we're learning from the internet?

Round 2: A series of occasional postings around the theme that patterns and ideas recur, or as Arlo Guthrie said in Alice's Restaurant, "come around again on the gee-tar."

tags:   | comments: 5   | Sphere It

Previous  |  Next

0 TrackBacks

TrackBack URL for this entry:

Comments: 5

  Michael Bernstein [08.19.06 08:37 AM]


Even as the world economy is busily becoming more interoperable and integrated, we are creating new economic spaces in virtual worlds, many of which are governed in ways intended to keep them isolated from the 'outside' economy.

Of course, any system that successfully keeps itself a silo will eventually die off, except for those that operate on a non-monetary basis.

The main question a few years ago was whether these virtual economies would first start interoperating with each other (ala GURPS) before integrating with the 'outside' economy, but that question has now largely been settled as 'no' (ie. you cannot directly exchange Lindens for Plats).

  Fred Zimmeramn [08.19.06 08:44 AM]

d !%@: A Directory of Electronic Mail Addressing and Networks

-- I had that book! It was fun.

  David Mercer [08.19.06 12:28 PM]

Oh wow, I've wanted another copy of that book for years, ever since I lost mine somewhere along the way. When I worked at an ISP in the mid-late 90s, actually knowing bang-path uucp addressing, and how to mix it with smtp-style @ addressing came in handy for one of our clients: they were doing anthropological field work in Afghanistan, and at the time the only email access in the country was through a BBS in that had a wireless link via uucp to a system in Pakistan. Madness, but I was able to get them sending email.

I know that nostalgia for that era is mostly elitism on my part, but man, the 'Net was a much more intellectually stimulating place then. Or at least it took a lot less effort to find intellectual stimulation of whatever sort you fancied, the signal/noise ratio being so much better then.

  Search Engines WEB * [08.19.06 07:17 PM]

One can not help but wonder - if the creators of all of these disparate networks had ORIGINALLY coalesced to create standards and contribute technical input, how much more advanced and secure would Network Hi-tech have been by now.

It is too late to start from scratch and rebuild, but any additions to the INTER-NET is simply demanding more additional middle layers to reinterpret, integrate and communicate.

  Michael Bernstein [08.20.06 07:57 PM]

Search Engine Web *, I actually suspect that had a deliberate interoperability effort been started sooner, instead of TCP/IP growing in the shadows and eventually taking over in a bloddless coup, we would have ended up with a much smarter (and hence worse) network than we have today.

One way of looking at this is that while smarter routers can make a network more efficient (and economic/profitable) in the short term, but at a cost of more brittle, less flexible network, with higher barriers to entry to connect as a node (which slows overall growth).

It is extremely hard to quantify the cost of lost opportunities in the form of forestalled unanticipated serendipitous innovations, so it's probable that we mostly wouldn't even be aware of the ways things could have been better.

The resulting network might have been 'good enough' to stop TCP/IP from winning, but it would have been worse in every important way than what we have today.

Arguably, that is exactly the sort of 'standardization' the telcos are trying to push through now, except we now *do* have a pretty good idea of what we would lose.

Post A Comment:

 (please be patient, comments may take awhile to post)

Type the characters you see in the picture above.