Metcalfe’s Law and the Crews Corollary

by on July 30, 2007 · 10 comments

I agree with Tim that open networks are great and likely preferable in most situations, but to say that open networks simply “tend to be better than closed networks” doesn’t make sense.

This is akin to saying that copper is more efficient than iron. This begs the question. More efficient at what? Copper is more efficient than iron in some applications like conduction of electricity, but it’s a much less efficient armor plating. Ends dictate the standard by which we judge efficiency, otherwise efficiency is meaningless.

That said, not all networks are built for the same ends. While the Internet is an undisputed engine of growth and innovation, it’s not the only model that EVER makes sense. Closed or limited networks can also have value because Metcalfe’s Law–which states that a network’s utility increases in proportion to the square of the number of members–is not the only factor in determining network worth, despite being a very strong factor.


Wayne Crews, my policy boss here at CEI, has proposed that Metcalfe’s Law needs a bit of an addendum. Wayne points out that there is an underlying assumption in this law that is crucial to its holding true. The “network effects” associated with Metcalfe’s Law will diminish if a sufficient number of users don’t have an interest in the health and security of a network. In fact, if members are added that are overtly hostile and seek to attack a network–exploiting features of its open architecture–then the value of that network decreases. Openness, therefore, is only valuable so long as the vast majority of users have an interest in the health of a network and the owner of that network can act to enforce policies that maximize network value.

I’m in favor of openness–it’s great! Most implementations of networks will benefit from a fair amount of openness, but it’s not the only way to make a network that will be of value.

I also agree that “it’s a mistake for libertarians to hang our opposition to government regulation of networks on the contention that closed networks are better than open ones.” But it’s also quite silly for the opposition to say that only open networks should exist. If it’s true that open networks are much better than closed, partly-closed, mostly open, or slightly-limited networks, then the market will bear that out and banning them outright would be unnecessary.

Robert Kahn, inventor of TCP/IP, has said publicly that he is “totally opposed to mandating that nothing interesting can happen inside the net.” I agree, network architecture differences should be a part of network competition.

  • http://www.techliberation.com/ Tim Lee

    Wayne Crews, my policy boss here at CEI, has proposed that Metcalfe’s Law needs a bit of an addendum. Wayne points out that there is an underlying assumption in this law that is crucial to its holding true. The “network effects” associated with Metcalfe’s Law will diminish if a sufficient number of users don’t have an interest in the health and security of a network. In fact, if members are added that are overtly hostile and seek to attack a network—exploiting features of its open architecture—then the value of that network decreases.

    Well sure, but this doesn’t contradict Metcalfe’s law unless you assume that the proportion (not just the number) of bad users grows with the size of the network. If the proportion of bad users is relatively constant (which seems like a reasonable assumption) then adding additional users will still increase the per-user value of the network, albeit not as quickly as a network with no bad users.

    Also, keep in mind that in most cases, security is not a characteristic of a network but of individual nodes or links in the network. The Internet has no owner, and therefore no one with the ability to police the security of the network as a whole. But each computer user, server administrator, ISP, or backbone provider has every incentive to police security on the part of the Internet they control.

    The Internet’s security problems are primarily a function of its size and the diversity of applications it supports. I see no reason to think that if the Internet were replaced by a proprietary network with a single owner that it would be any more secure. Quite the contrary, if the Internet had a single owner, that owner would be overwhelmed with the logistical challenges of imposing a single, top-down security model across the network. The principles of decentralization and spontaneous order apply to network architecture as much as they do to economics.

  • http://cei.org Cord Blomquist

    Point by Point:

    You’re right in saying that the proportion of bad users is likely to be relatively constant were the scale of the network to have no bearing on the incentives of would-be hackers. However, the bigger the network, the greater the reward if a virus is a success, for example. This may be why there are next to no virus attacks on Macs–there isn’t a sufficient mass of users to reach an incentive tipping point.

    Security on the internet is largely relegated to the individual machine, but the network has something to do with it! That’s where, like you say, ISPs come in. My point is that if ISP are forced by law leave all security at the edges of the network, the network will suffer overall. A smart network with policies that can reduce the effects of the black-hats in the user population, will be more valuable.

    Finally, I’m not arguing for the internet as a whole to be owned by one entity. A network of networks makes sense for reasons of decentralization and spontaneous order like you said. Even so, you can apply the Crew Corolary to networks within the internet.

  • http://www.techliberation.com/ Tim Lee

    Wayne Crews, my policy boss here at CEI, has proposed that Metcalfe’s Law needs a bit of an addendum. Wayne points out that there is an underlying assumption in this law that is crucial to its holding true. The “network effects” associated with Metcalfe’s Law will diminish if a sufficient number of users don’t have an interest in the health and security of a network. In fact, if members are added that are overtly hostile and seek to attack a network—exploiting features of its open architecture—then the value of that network decreases.

    Well sure, but this doesn’t contradict Metcalfe’s law unless you assume that the proportion (not just the number) of bad users grows with the size of the network. If the proportion of bad users is relatively constant (which seems like a reasonable assumption) then adding additional users will still increase the per-user value of the network, albeit not as quickly as a network with no bad users.

    Also, keep in mind that in most cases, security is not a characteristic of a network but of individual nodes or links in the network. The Internet has no owner, and therefore no one with the ability to police the security of the network as a whole. But each computer user, server administrator, ISP, or backbone provider has every incentive to police security on the part of the Internet they control.

    The Internet’s security problems are primarily a function of its size and the diversity of applications it supports. I see no reason to think that if the Internet were replaced by a proprietary network with a single owner that it would be any more secure. Quite the contrary, if the Internet had a single owner, that owner would be overwhelmed with the logistical challenges of imposing a single, top-down security model across the network. The principles of decentralization and spontaneous order apply to network architecture as much as they do to economics.

  • http://www.cordblomquist.com cordblomquist

    Point by Point:

    You’re right in saying that the proportion of bad users is likely to be relatively constant were the scale of the network to have no bearing on the incentives of would-be hackers. However, the bigger the network, the greater the reward if a virus is a success, for example. This may be why there are next to no virus attacks on Macs–there isn’t a sufficient mass of users to reach an incentive tipping point.

    Security on the internet is largely relegated to the individual machine, but the network has something to do with it! That’s where, like you say, ISPs come in. My point is that if ISP are forced by law leave all security at the edges of the network, the network will suffer overall. A smart network with policies that can reduce the effects of the black-hats in the user population, will be more valuable.

    Finally, I’m not arguing for the internet as a whole to be owned by one entity. A network of networks makes sense for reasons of decentralization and spontaneous order like you said. Even so, you can apply the Crew Corolary to networks within the internet.

  • http://www.techliberation.com/ Tim Lee

    However, the bigger the network, the greater the reward if a virus is a success, for example. This may be why there are next to no virus attacks on Macs–there isn’t a sufficient mass of users to reach an incentive tipping point.

    Well sure, but I still think it’s hard to deny that network effects are a net positive for Windows. Despite the viruses, people are still overwhelmingly choosing to use Windows, largely because of the greater availability of hardware and software. If your claim were correct, we’d reach a tipping point where the virus problem was so severe that people started leaving Windows in droves simply to get away from virus problems. (Yes, Apple is gaining market share, but very few of the switchers are motivated primarily by virus problems).

    That’s where, like you say, ISPs come in. My point is that if ISP are forced by law leave all security at the edges of the network, the network will suffer overall. A smart network with policies that can reduce the effects of the black-hats in the user population, will be more valuable.

    Absolutely. But then your beef isn’t with “openness” as such, but with stupid government regulation. Adding a firewall or a spam filter to an open network doesn’t make it a closed network. And, for that matter, Snowe-Dorgan wouldn’t prevent ISPs from using firewalls or spam filters. There are a lot of problems with Snowe-Dorgan, but this isn’t one of them.

  • http://www.techliberation.com/ Tim Lee

    However, the bigger the network, the greater the reward if a virus is a success, for example. This may be why there are next to no virus attacks on Macs–there isn’t a sufficient mass of users to reach an incentive tipping point.

    Well sure, but I still think it’s hard to deny that network effects are a net positive for Windows. Despite the viruses, people are still overwhelmingly choosing to use Windows, largely because of the greater availability of hardware and software. If your claim were correct, we’d reach a tipping point where the virus problem was so severe that people started leaving Windows in droves simply to get away from virus problems. (Yes, Apple is gaining market share, but very few of the switchers are motivated primarily by virus problems).

    That’s where, like you say, ISPs come in. My point is that if ISP are forced by law leave all security at the edges of the network, the network will suffer overall. A smart network with policies that can reduce the effects of the black-hats in the user population, will be more valuable.

    Absolutely. But then your beef isn’t with “openness” as such, but with stupid government regulation. Adding a firewall or a spam filter to an open network doesn’t make it a closed network. And, for that matter, Snowe-Dorgan wouldn’t prevent ISPs from using firewalls or spam filters. There are a lot of problems with Snowe-Dorgan, but this isn’t one of them.

  • http://cordblomquist.com Cord Blomquist

    Tim, I think we really agree on the underlying issue, which is that no matter what the worth of closed or open networks, neither should be given preference by law and neither should be banned outright.

    Ultimately, our thoughts on the value of networks don’t matter. Free markets should be “the decider,” not guys like us chattering, though I’m happy to continue the chatter.

  • http://www.cordblomquist.com cordblomquist

    Tim, I think we really agree on the underlying issue, which is that no matter what the worth of closed or open networks, neither should be given preference by law and neither should be banned outright.

    Ultimately, our thoughts on the value of networks don’t matter. Free markets should be “the decider,” not guys like us chattering, though I’m happy to continue the chatter.

  • http://linuxworld.com/community/ Don Marti

    There is one kind of network user that needs to be kept on open networks by law — the government itself.

    If the government puts the equivalent of a .gov site on a closed network, or requires some communication from citizens to be in a format that requires proprietary software to create, then the government is effectively adopting the network ToS or the software EULA as regulations. By requiring the government itself to use open networks and software, you’re reducing the total regulatory burden when you add in all the people who have a reason to communicate with the government.

    So open networks can’t go away entirely — either the government has to be in the business of building them, or whatever regulatory framework is adopted needs to require open networks for citizen-government communications.

  • http://linuxworld.com/community/ Don Marti

    There is one kind of network user that needs to be kept on open networks by law — the government itself.

    If the government puts the equivalent of a .gov site on a closed network, or requires some communication from citizens to be in a format that requires proprietary software to create, then the government is effectively adopting the network ToS or the software EULA as regulations. By requiring the government itself to use open networks and software, you’re reducing the total regulatory burden when you add in all the people who have a reason to communicate with the government.

    So open networks can’t go away entirely — either the government has to be in the business of building them, or whatever regulatory framework is adopted needs to require open networks for citizen-government communications.

Previous post:

Next post: