On the invention of the internet

Update: A more active thread on how this relates to Goodmail and other attempts at sender-pays traffic

There is much talk these days of "who invented the internet?" Most of the talk is done wearing a network engineer's hat, defining the internet in terms of routing IP datatgrams, and TCP. Some relates to the end to end principle with a stupid network in the middle and smart endpoints. These two are valid and vital contributions, and recognition for those who built them is important.

But that's not what the public thinks of when it hears "the internet." They think of the collection of cool applications they use to interact with other people and distant computers. Web sites and mailing lists and newsgroups and filesharing and VoIP and downloading and chat and much more. Why did these spring into being in this way rather than on other networks?

I believe a large and necessary ingredient for "the internet" wasn't a technological invention at all, but a billing system. The internet is based on what I call the "internet cost contract." That contract says that each person pays for their own pipe to the center, and we don't account for the individual traffic.

"I pay for my half, you pay for yours."

While the end-to-end design allowed innovation and experimentation, the billing design really made it possible. In the early days of the internet, people dreamed up all sorts of bizarre applications, some serious, some entirely frivolous. They put them out there and people played with them and the most interesting thrived.

Many other networks had users paying not by the pipe, but based on traffic. In that world, had you decided to host a mailing list, or famously put a webcam up in front of your company fishtank, the next day the company beancounter would have called you into the office to ask why the company got a big bandwidth bill in order to show off the fishtank. The webcam -- or FTP site or mailing list -- would have been shut down immediately, and for perfectly valid reasons.

Pay-based-on-usage demands that applications be financially justifiable to live. Pay-per-pipe allowed mailing lists, ftp sites, usenet, archie, gopher and the web to explode.It's not black and white of course. Some apps are financially justifiable and could thrive in any environment, and perhaps might have thrived more. Pay-per-use environments with complex arrangements for caller-pays and recipient-pays ("collect") billing can enable even more, though this was not common. Imagine trying to host a big mailing list on MCI mail, where the sender paid for each mail?

Even so, pay-by-the-hour environments like Compuserve did manage to do well when their cost was low enough that people weren't watching the clock so much. But only in specialized areas, and their growth was nothing like the internet. In time, the last of them (AOL) switched to flat rate.

Now the problem with the description I've just given of the enabler of the internet is that it has no specific parents to name as inventor, the way packet switching might. Creating systems without accounting is common in research projects, which is how the net began. The novelty was keeping that no-accounting model as the net expanded out into the public sphere in the 1980s. Some of the people responsible for this include NSFNet (Steve Wolfe) and the CSNet folks who pushed for pay-by-the-pipe accounting as they connected to the ARPANet.

There were also folks who just ate the bills from pay-by-usage networks (like the PSTN) to carry USENET and FidoNet and give the illusion of pay-by-pipe (or free) to their users.

So who is responsible for this core invention that made the internet happen? The pricing model has many parents, and none. The first network connections were built that way because it was much simpler to do it that way. Having talked to those who were there at the time, like my friend Dave Farber, they didn't realize how the pricing model would foster innovation. Being in an academic (or academic-like) environment it was simply the case that the hard task of making a complex billing system wasn't even remotely on the priority list. In contrast, the CCITT/ITU teams that defined X.25 were telco people, who considered billing and settlement from the start. The X.25 networks offered inter-network, international packet switching service on a commercial basis a decade before the internet was commercialized, but they fell before the internet.

Of course, some will argue that the lack of a usage accounting structure had negative effects. "900" style service or micropayments might have enabled some real content businesses rather than pushing everybody to the advertising model. And some abuses, such as spam and DDOS would be easier to fight in a pay-by-usage network. (Indeed, spam is making many call for the abandonment of the internet cost contract, though I think they don't know what they are really asking to destroy.) I remember throughout the early days of commercialization how many people insisted that this crazy unbilled system could never work, that settlement payments were the only answer.

In the long run, it is the per-pipe accounting that gave us the internet and we would be fools to destroy it.


It's easier to trace the roots of this if you look at paying not "for your half" but "for your pipe to the rest of the network".

UUCP mail and USENET newsgroups often worked this way. A site already on the network would allow others to phone it (at their own expense) and thereby join the network. The original nonprofit UUNET, a telephone UUCP/USENET service, was designed to reduce these phone costs, particularly for people who had to call long-distance, by aggregating the calls into one company, getting a volume discount, and charging at a rate designed to recover the costs. It worked.

When three companies started the second Bay Area regional ISP, The Little Garden, in 1990, we split down our $1000/mo UUNET Internet bill evenly, plus, each of us also paid the cost of our own 56K leased-line (or dialup modem-line) that tied us into the nearest point on the network. After connecting about thirty individuals and companies, things got more formal, with these costs more buried in generic pricing. TLG became the backbone of many dozens of little ISPs in N. Calif.

Add new comment