How the Net got Neutral

Scott Bradner remarks that Dennis Jenning reminded him that the NSFNet project was approved 21 years ago. In case you've forgotten or weren't on the net back then, NSFNet was a fast (T1! Really fast for back then!) backbone network meant for primarily academic use, and that also served as the primary backbone of the Internet. This was in 1985, remember, six years before the first commercial long-haul ISPs (UUNET and PSINet).

By one measure the Internet has reached the age of majority in Washington, D.C. - one of the places that seems to have the worst understanding of what the Internet was, is and can be. The technology trickle that became the Internet started with research into packet-based networks in the early to mid 1960s by Len Kleinrock, Larry Roberts, Paul Baron and others (Google can help you find lots of information on these folks).

Key decisions that enabled the 'Net By Scott Bradner, 'Net Insider, Network World, 10/02/06

Scott lists ten design decisions that let the Internet be what it is today.

Here they are, in my paraphrase:

• use existing networks;
• packet switching;
• split reliable TCP from routing IP;
• U.S. funds UC Berkeley reference implementation of TCP/IP;
• CSNet brings ARPANET mail access to many universities;
• NSFNet requires TCP/IPon NSFNet;
• ISO declines to standardize TCP/IP;
• NSF blocks commercial use of NSFNet, leading to commercial ISPs;
• U.S. government does not regulate the Internet.

I can't argue with any of those, but I think there were others:

• Requirement of TCP/IP to remain connected to the ARPANET in 1983.
• Development of the distributed and decentralized Domain Name System (DNS) to replace the old ARPANET host table, so that there was no need to apply to a single centralized authority to connect.
• UUCP, which brought electronic mail to many companies that didn't have ARPANET or CSNet access.
• FidoNet, which brought electronic mail to many people who didn't have UUCP or ARPANET or CSNet access.
• BITNET and EARN, which brought networked electronic mail to big iron (IBM and Digital mainframes).
• USENIX provides initial funding for UUNET, in 1987.
• Early commercial ISPs eschew settlements, ca. 1991. Volume charging had been one of the main things that killed European-style X.25 and ISO/OSI networks, and settlements were a characteristic feature of telephone networks, which were in many countries run by the same organizations that ran the X.25 and ISO/OSI networks. With volume charging and settlements, we never would have seen anonymous FTP, WWW, blogs, BitTorrent, YouTube, etc.

On the U.S. government regulation front, Scott notes that Dennis Jennings' requirement of TCP/IP on NSFNet was a mandate, while most of the other items weren't. DCA's requirement of TCP/IP on ARPANet in 1983 was a similar mandate. Those were actually useful because they were for networks directly funded by that government, and they promoted new technology.

Another mandate was GOSIP, the requirement that any operating system software sold to the U.S. government had to include the ISO/OSI protocols. That one was not so useful, since it wasted a lot of vendor time in implementing a protocol suite that had already lost the technological and user wars, even though many European countries and Japan and others had actually required use of ISO/OSI. Eventually GOSIP was terminated.

Mandating specific technological solutions is almost always a bad idea in networking. Which is one reason why any network neutrality laws have to be carefully worded. Another reason is to avoid prohibiting any more business models than necessary, because you never know which one might produce the most value for everyone, not just for the telcos.

However, as Scott notes, one of the reasons packet switching was a big win was that it didn't require a carrier to be involved in setting up communications. Letting the carriers determine what communications can be set up would be a huge step backwards, and that is one of the main reasons we need network neutrality.

Scott remarks of Washington:

The feeling there now is that the Internet is far too regulation-free (and maybe too good at innovating). Congress and the FCC are fighting to fix this perceived problem.

I have no idea if we will be able to look at the Internet 10 years from now and see anything we would recognize as the Internet. Many in Washington seem to hope not.

Sen. Stevens has already said he plans to pass his telecom. bill in the next Congress. We'll see.

-jsq