On August 23 ( while I was in China) a list member Lee S. Drybrugh wrote in jest:

I happened to bump into Peter Cochrane stating: “The good news is—bandwidth is free—and we have an infinite supply.”

Can anyone explain this infinite bandwidth as I think I am getting ripped off by my ISP if this is true?

Craig Partridge then offered what I think is a very good commentary of a difficult question where the answer depends very much on context.

Craig wrote: It all depends on what you are measuring and where.

If you look purely at how many bits we can (in theory) push through a fiber and look at the available fiber base in many first world countries you’d discover that, in theory, we’re so underutilizing the fibers that we could, with a small stretch, treat available bandwidth as infinite.

Similarly, if you stood just about anywhere in the world and measured how much of the wireless spectrum was in use at that spot, you would find so little of the spectrum used, you could say the available bandwidth was infinite. Cue the software defined radio theme music :-).

Satellites are more cautionary—in theory there’s almost infinite bandwidth available optically between satellites (assuming we can get freespace optics between satellites to work), but the combination of a limited number of geostationary orbits available and the challenges of communication with the ground means capacity is rather limited. Infinite bandwidth in space isn’t useful unless you can, somewhere, get those bits down to Earth. (Lower orbit satellites remix the tradeoffs but still have limits).

If you look up from the communications media to things such as routers and switches and network operation center costs and such like, it is clear that bandwidth is *not* infinite. From 1969 to 1997 or so, we were constantly challenged to build communications equipment that had enough capacity to keep pace with demand and improvements in fiber optics—we’ve been in a lull recently, but it may not last. To exaggerate this point, it is conceivable we can create links so fast we cannot interconnect them—we were flirting with that disaster in the mid-1990s. And if you cannot build the interconnection devices, the whole “infinite” argument collapses as we certainly don’t have infinite point to point capacity between all interested parties.

Network operations costs per bit moved have declined sharply over the past 40 years but my sense (quite possibly wrong) is that costs have declined only slightly in the past decade. Exactly how this effects the “bandwidth is infinite” argument is not entirely clear to me—except that we can conceivably build networks with more bandwidth than we can afford to operate.

Update (Sep 15, 2007): Permission granted to include Lee S. Drybrugh’s name in the first sentence of the post.