If you’re like most of the rest of us, the speed at which you upload is way slower than the speed at which you download. That makes sense if you think of the Internet as a publishing medium, with most of its participants as recipients. It makes much less sense if you think of the Internet as a participatory medium.

So, I canvassed some friends and colleagues who know a lot about this stuff — I’m leaving them unnamed so my mistakes won’t besmirch them — about whether this asymmetry is baked into the technology or whether it’s a matter of choice and policy. The answer turns out to be complex. But the bottom line is that we download faster than we upload because of economics, not physics. The economics are those of telephone and cable companies. Their business models assumed the value of the Net came from the content they deliver to us. The business models got written into a physical infrastructure that favors downloads over uploads. So, here we are.

It could have been otherwise. If only the access providers hadn’t built a business and an infrastructure on the assumption that we really just want to lean back, relax, and scarf down gigabits of tasty content.

Physics of course sets the boundaries for any business model. If you’re uploading and downloading through the same “pipe,” increasing upload speeds requires decreasing download speeds. To oversimplify incredibly complex technology, it’s like having a single tunnel for one-way traffic: If you now make it two-way (instead of, say, putting in another tunnel), you’ve reduced the one-way capacity. When the Internet started almost all of us downloaded far more than we uploaded — simply visiting a Web page requires downloading it. Asymmetry reflected our typical usage pattern. Most of us still pull down more than we put up, but with the rise of P2P networks, voice over IP, etc., we’re doing much more uploading than we used to. And if the access providers made static IP addresses more affordable, we could host our own Web sites and be uploaders to everyone who visits.

Nevertheless, users generally aren’t picking up their pitchforks and demanding higher upload speeds … perhaps because when we can’t upload very quickly, there’s little market for services that require a lot of uploading, so people don’t build upload-heavy services, so we don’t see good reasons to demand faster uploading…

The answer to the question of asymmetry gets more complicated when you look at the various ways of connecting:

When copper cables were first used for DSL, uploads were assigned to a low frequency because there is less attenuation over distance (you hear the bass of a distance band far more clearly than the treble), but lower frequencies carry less information. At the time, the lower upstream bandwidth didn’t seem important. In fact, Asymmetric DSL (ADSL) was based on the assumption that the only uploading of data we’d be doing would be for “interactive TV.” Symmetric DSL (SDSL) does exist, but the trade-off is that you get lower download speeds. Wrote one of the the people I asked: “On a single copper line I’ve never heard of a commercial offer above 2Mb/s symmetric.” Meanwhile, he points out, an ADSL2 line can run at 15Mbps down and 1Mbps up. This trade-off really has to do with a business deal, he says: SDSL gives guaranteed speeds to its users, so it has to be safely low, whereas ADSL’s higher speeds are not guaranteed; ADLS’s high advertised speed represents a best case that few may ever actually experience … and that none can sue over.

In the little discussion that emerged around my question, there was disagreement about whether coaxial cable has physical limitations that make symmetry difficult. One person contended that with coax, many end users are contending for access, limiting the available bandwidth. Also, having lots of users pushing things up multiplies the amount of noise, which again limits the effective bandwidth. Others claimed that coax itself has no problem with symmetry, but the way it has been deployed “has a lot of asymmetry baked in because the amplifiers and taps the cable passes through in the system were built with the assumption that traffic would only be going one way.” Upstream was assigned a low frequency, for the same reasons as with copper telephone wires, but the allocation could be changed if the provider wanted to dig up cables and invest in new hardware. Conclusion for coax: “Physics and historical choices have biased the access network towards asymmetrical bandwidth.”

For optical fiber, the group seems to agree that the asymmetry is a packaging decision that is now being baked into deployments. Most of the incumbents are deploying fiber using GPON (Gigabit Passive Optical Network), which enables a single strand of fiber to serve multiple users; the same data stream is delivered to each user, encrypted so that only the relevant signals are decoded by each. GPON dynamically divvies up upload and download, with the standard set to 2x download over upload. Verizon’s FIOS fiber service is more asymmetric than that because that’s how Verizon wants it.

So, the answer to the overall question is that physics requires an apportioning of bandwidth between up and down, but economics can here route around physics: If a provider were committed to providing symmetry without lowering download speeds, it could have (for example) put in a second line when it put in the first. The trade-offs were made by the access providers, largely based on business concerns. For example, one reason we have asymmetric fiber may be that the providers have been charging such high premiums for symmetric access that they don’t want to lose that extra money, even as the technology makes symmetry more plausible.

So, as was said by David Isenberg (who gave me permission to credit him): “Asymmetry is a belief system. The purveyors of connections looked at the Internet and saw TV, then acted according to what they saw.” That’s the infrastructure they built. That’s the infrastructure we’re currently stuck with. If we can get unstuck, we might at long last get ourselves some symmetry…or have broad enough broadband that symmetry doesn’t even matter.

[Note: A shorter version of this piece is up at HuffingtonPost.]