December 27, 2013

Why the Internet is, by Definition, Ungovernable


To the Internet community, it could be the end of the world as we know it, or it could be much ado about nothing.  International organizations such as the UN’s International Telecommunications Union have been debating “internet governance.” Some individuals, and some governments, are under the impression that the Internet is run by the United States government, and that it is too important to be left to one country’s control.  The 2013 Snowden revelations about mass surveillance by the NSA did little to allay that distrust of the United States, though there is no real connection.  Thus there is considerable interest, especially outside of the US, in creating a new international governance structure for the Internet, one not under American control.

The problem with this approach is that it totally misunderstands what the Internet is and how it works.  The United States does not govern the Internet.  Nor does ICANN, the California-based not-for-profit that administers the key directories that Internet users depend upon.  No one governs it, and that’s precisely why it works.

Just what is the Internet?

The big-I Internet is a prototype implementation of the (small-i) internet concept.  There can be other internets.  They don’t have to use the TCP/IP protocol, which is almost 40 years old.  They don’t have to be worldwide or open to all comers.  But the big-I Internet, using IP version 4 (which only dates back to 1978) will remain the biggest one for the foreseeable future.  (IP version 6 is going nowhere fast, has been about two years away from replacing v4 for the past decade or more, and isn’t getting any closer). So while there is really one Internet that counts today, it is foolish to think of it as the final stage of networking.  Both the Internet, and internetworking, are still evolving.

Several years ago, I proposed this definition of internet:

Internet (n.) A voluntary agreement among network operators to exchange traffic for their mutual benefit.

This is a definition, not a description.  It sets the broad boundaries of what the word “internet” refers to. It also happens to nicely describe how the Internet evolved over the past two decades. 

An internet is essentially a network that operates within the payload of multiple autonomous networks.  While some early Internet documents refer to it as a catenet, that term is inappropriate.  It is critical to distinguish between a catenet and an internet.  A catenet is functionally a single network that has multiple owners, but operating under common policies.  The different networks thus pass traffic to one another, “beads on a string.”  There is no functional layering involved.  Everyone is of course familiar with the granddaddy of catenets, the public switched telephone network.

An internet takes advantage of a functional boundary between the underlying networks and the internet itself.  Traffic is relayed between nodes on the internet using whatever lower layer network is available.  In today’s Internet, IP is relayed between routers, but the underlying “network” links could be local area Ethernet networks, metropolitan-area Carrier Ethernet, WiFi, LTE, leased lines, MPLS, ATM, Frame Relay, etc.  That’s what makes the concept powerful, and it’s not a new idea.  Vint Cerf and Bob Kahn invoked it in 1978, when they wrote IEN54, the original description of IP version 4.  It included this picture:

Note that the term “Local Network” did not mean “LAN” at the time; it referred to whatever linked the IP nodes.  In 1978, many countries were developing their own packet-switched networks based on the then-new X.25 standard.  X.25 was a rather loose standard with many options that led to network-specific “profiles”.  This complexity was hidden by IP, since it was encapsulated by a network interface and each hop could use a different network. X.25 specified a network; IP was used by an internetwork.

An internet did not require central management.  TCP provided “end to end” error detection and recovery, not trusting the underlying neworks.  This idea actually predates TCP; Louis Pouzin came up with it when he developed CYCLADES, the first connectionless network, in France, beginning in 1972.  He recognized the power of layered protocols.

Internets are voluntary

Before 1992, “The Internet” was indeed a creature of the United States government.  It grew out of the ARPANET, owned by the Department of Defense Advanced Research Projects Agency.  That demonstrated packet switching in 1969, fully adopted TCP/IP in 1983 and provided connectivity between government researchers until being wound down in 1987.  By 1992, the most important Internet backbone was the NSFnet, funded by the National Science Foundation, another federal agency.  Regional member networks connected universities, who owned them, and some corporations.  But it was not open to the public.  There was no e-commerce – commercial use was banned by the NSF Acceptable Use Policy (AUP). By 1993, the AUP had ended, NSFnet was being phased out, and several commercial operators were taking up the slack, in some cases buying up NSFnet regional networks. 

The business model of the commercial Internet fell into place rapidly. It was not subject to government regulation, though the federal government still supervised some coordinating activities, such as the address and name registries.  In the United States, the Computer II framework was in effect, defining computer services as “enhanced”, and requiring telecommunications providers to offer “basic” connectivity to them.  That vital split, between content (Internet) and carriage (telecom), prevented the slow, complex telecom regulatory process from preventing the Internet from taking off commercially. The Internet, then, is not a thing, and not even a network that can be governed by regulated tariffs, but a set of mostly-bilateral agreements.  That they are voluntary means that they are not subject to external governance. This flexibility helps it overcome the brittleness inherent in its creaky old protocols.

The Internet is a thus a phenomenon. It was empowered, not created, by regulatory protection.  And misguided attempts to regulate or govern it directly, based on a misunderstanding of the phenomenon, can only break it.  The underlying networks can and should be regulated, of course.  The current US regulatory model, which deregulates telecom facilities when they are carrying Internet as payload, totally misses the point.  Internets are complementary to, but different from, and in fact dependent upon, regulated public networks.

Internet providers exchange traffic for their mutual benefit.  This is key to the business model.  It’s like a real business, and unlike the artificial world of telecom.  A bigger network is worth more than a smaller one, due to the network effect.  Thus a smaller network gains more by exchanging traffic with a bigger one.  Therefore the mutual benefit of interconnecting networks may involve the exchange of money.  It is a free market system where size matters. 

This arrangement only works because, after the NSFnet exit, there was no one dominant backbone ISP.  Instead, a voluntary system of peering evolved.  Big networks peer with each other for no charge.  Networks also provide transit to each other.  So if you don’t like the deal you get with one network, you can still reach it via a third party.  That is quite a contrast to the PSTN model, where prices are regulated, size doesn’t matter, and transit providers usually don’t have responsibility for end to end billing.

There’s also no bright line between “customer” and “peer.” PSTN carriers interconnect with each other, while they sell service to non-carrier customers.  Just being a carrier, however small, makes a difference.  In contrast, there is no licensing of ISPs, no formal certification that entitles one to peering.  It’s all negotiable.

And being voluntary means that there is no obligation to carry anything. An internet is not a common carrier.  Thus “network neutrality” cannot be imposed upon an internet, or by definition it is no longer an internet, just another regulated public network.  It doesn’t matter if it is an IP network. Protocol does not define an internet!  The telephone network is using IP internally but it is not an internet.  While there are misguided attempts to impose common carriage upon the Internet today, such a move would be highly destructive.

Governments can’t override users

The International Telecommunications Union (ITU) has, since the 19th century, coordinated government-regulated public networks.  It thus makes rules that are meant to apply to all PSTN carriers, who operate under tariffs.  But the Internet is simply a set of voluntary agreements.  It’s contracts, not tariffs, and thus subject to ordinary contract law, not telecom regulation.

Thus a lot of activities that we take for granted are merely voluntary choices.  If the ITU or some other agency tried to hijack the DNS root, for instance, alternatives could spring up.  Alternate routes, like open-route.eu, already exist.

We already see non-free counties setting up national “internets.” These do not have full connectivity to the rest of the world.  There are also ISPs that choose to restrict connectivity based on their customers’ religious preferences.  It is their choice.  We do not have to approve of censorship, but it’s out of our control.

So if governments tried to take control of the Internet and manage it in a way that users didn’t like, the likely result would be the creation of new internets that rode atop the regulated ones, again not trusting them, routing around obstructions.  The Internet already hosts a number of darknets, secret internets that prove the possibility of hiding an internet inside another network.

Internet “governance” today, including the role of ICANN, is thus merely consultative. It’s not like the PSTN where the name space (telephone numbers) really is government owned. The process works so long as it is necessary and useful.  Public IP addresses need to be coordinated within an IP-based internet in order to avoid conflict.  But “pirate” addresses already exist in today’s Internet, and other internets could use alternative address schemes, or other protocols, via translating gateways. People, or their ISPs, merely choose to use the ICANN root, via name servers that point to it, but individuals can point their DNS resolvers at servers other than their ISPs’.

Thus, by definition, the Internet is not governable.  ICANN is a consultative body with no legal authority, just its own credibility.  While its decisions have at times been controversial, it has generally done its job well enough, and thus it remains the coordinator of choice.  Any attempt to create an actual governance scheme is thus likely to fail.




Edited by Cassandra Tucker



Related Tags

Networking       
       

blog comments powered by Disqus

More in TechZone360