Mischa Beitz

Akamai's Edgesuite product does not break network neutrality.

Apparently critics of network neutrality are still trotting out the argument that Akamai's Edgesuite product breaks network neutrality (most recently, Christopher Yoo raises the claim in a slashdotted interview; (NOTE: Since I originally posted this blog, the article has been substantially edited. Nevertheless, while not as strongly made, the Akamai point is still there) though he originally made the claim much earlier - it's also been raised at times by the likes of Adam Thierer and others). The claim Yoo et al. make is dangerous (I believe) because it relies on an idea of "network neutrality" that seems very much like what some proponents favor, but very unlike what other proponents (including me) argue for. In other words, it's an argument that threatens to divide network neutrality proponents.

Akamai's Edgesuite product can be summarized quite simply. Akamai has a global network of servers which mirror internet content (and dynamic services, it's not just static content - they do some very cool stuff). Edgesuite hosting then uses DNS resolution to direct requests for an online service (HTTP, FTP, SMTP, etc) to one of Akamai's "nearby" (this could be in terms of geography, network distance or even internet traffic patterns - I don't know exactly what they employ, but like I said, they do some pretty cool stuff) servers rather than have the request hump it's way across the globe to find the client's server. By using Akamai's Edgesuite product an Edgesuite client's web-content would then load more quickly than someone who didn't use it, simply because it wouldn't have to "travel as far". Yoo et al. suggest this breaks network neutrality because it enables content or a service provided online to respond more quickly and reliably by purchasing Edgesuite, thus breaking the "level playing-field" that (according to some proponents) network neutrality is supposed to protect.

Yoo et al. (and even some network neutrality proponents) have an unfortunately misguided idea of how network neutrality ought to be understood. The basic idea is captured in a famous (and famously misunderstood) paper by Saltzer, Clark and Reed in which they present the end-to-end principle. The end-to-end principle basically says, if you CAN put functionality at the END of a communication rather than IN the communications medium, then you SHOULD put the functionality at the end. The idea is more or less synonymous with the idea of a "dumb network". What Yoo et al. (and the majority in the Brand-X decision - I still can't believe it) don't seem to understand is that DNS is an example of EXACTLY this principle; i.e DNS is the implementation of a service that sits at the END of the communication and is NOT a part of the communication medium. If you're having trouble understanding this (as Yoo and 6 out of 9 supreme court justices do), pull out your handy-dandy OSI network layers chart and figure out where to place DNS; hint: "applications layer". Network neutrality is NOT about (or at least ought not be about) regulating the applications layer, it's about ensuring the end-to-end principle so that companies like Akamai can innovate at the ends of the network. Far from being an example of breaking network neutrality, Akamai services like Edgesuite are (yet another) example of the kind of innovative application that a neutral network ENABLES!

Richard Whitt, Kevin Werbach and others have proposed regulatory reform based on the network layers model and, in general, it's the best basis for reform I've seen so far. Hopefully the new congress will take their suggestions seriously. I'm open to alternatives, but given the history of our (information and telecommunications) "services" based regime, I think embarking on another definitionally squishy "services" regime where IPTV services, VoIP services and Broadband services dissolve into a litigated mush of regulatory ambiguity is a particularly bad idea.

On that note . . . I hope to blog on FCC approval of the AT&T/BellSouth merger next.