A thoughtful reader with no connection to cable companies thought my analysis of DSL versus cable service (in re: SBC's Wi-Fi hotspot offering) was a little harsh: This reader has had great cable modem performance, meeting the promised speeds. But I thought I'd share with you my response. Update: Folks more knowledgeable than I in the workings of the cable world have taken me to task: apparently there's more in common now with DSL and digital cable head ends than a few years ago--my knowledge on this subject is obviously out of date. I've revised this post.
When you buy into a pooled business service like Frame Relay, you can opt for a Committed Information Rate (CIR), or a rate below which the bandwidth will never be unavailable, but above that rate, you're not promised and you may often get higher rates. At one point, I had a 1.5 Mbps frame connection with a 384 CIR. I usually saw peak performance. But I was paying for a pool of bandwidth shared (virtually) with other users from other companies.
I believed that cable still had this same difficulty. However, digital cable head ends, or the points at which entire systems are connected (usually by optical fiber) out to neighborhood cable nodes which lead to hundreds of homes by coax, has much higher capacity and much greater ability to expand than I realized.
With DSL, the effective limit is the size of the DSLAM or DSL aggregator/multiplexer in the C.O. or central office. From the C.O. in most areas that have DSL, there's probably fiber back to the Internet, or something sufficiently high-speed to provide a large enough pool at that end.
So the pool that DSL users share could be higher because of the one-to-one DSL modem to DSL card in the DSLAM relationship. The pool that cable modem users share is at the neighborhood cable node, which limits the maximum amount of bandwidth that the cable companies can offer across the entire neighborhood shared network.
Where I erred in my analysis in this post was thinking that the cost and complexity of cable companies increasing bandwidth more than moderately as demand increased would radically outstrip capacity. But, as one reader, Mike Ritter noted, "It's an open issue if their bandwidth is more expensive to provide than DSLAM bandwidth. My guess is not."
Kerry Williamson brought up an excellent counterpoint to my one-to-one DSL argument, too: "Cable companies are able to provide exactly the same level of service everywhere within their plant [wire service area]. That LOCAL area plant can have a radius of 100 miles (I know, I have designed and built several of them), and have no effect on the service, either data or video. The telephone company and DSL cannot do that. The further from the CO, slower the service. Not so with cable." The more I read about this issue, it's certainly cable's greatest single advantage over DSL's current few mile limit for high-speed performance.
Williamson notes that current technology has the ability to offer in the U.S. 10 streams or pools of about 30 Mbps downstream and 10 streams of 3 Mbps upstream without any real difficulty using the widely adopted DOCSIS 2.0 cable data protocol. Each stream is a 6 megahertz wide (replacing the spectrum used for a broadcast television channel), offering a pool as large or larger than non-fiber-based DSLAMs in telco offices. DOCSIS 2.0 can run over 40 Mbps in the U.S., but at those speeds has much greater susceptibility to noise. (European broadcast channels are wider, providing more spectrum that can be replaced in the cable system by data, increasing bandwidth per stream.)
Williamson also notes that a new revision to the cable modem standard, DOCSIS 3.0, could bring 200 Mbps per stream downstream into the home. Here's an article in a broadband publication that details DOCSIS 3.0. Upstream speeds would dramatically increase, with 100 Mbps available per pool in the best-case scenario.
Byoung Jo Kim responded to the issues of whether there's more expansion left in DSL versus cable with this: "For DSL, the high rates achieved in South Korea and Japan are mainly due to their shorter lines and higher line qualities from denser and more recent deployments of twisted pair copper wires. In the US with the longer and older lines, it will be very difficult/expensive to achieve such rates just by putting in new line cards at central offices. Thus, even Bells are looking into wireless for reaching far-away houses that HFC [hybrid fiber coax] reaches easily, although the seriousness is questionable."
Kim points me to this reference: "A view of fiber to the home economics" by Frigo, et al, in Communications Magazine, IEEE (Aug. 2004, pages S16-S23, Vol. 42, Issue 8). (Only an abstract is available at no cost to non-IEEE members.)
Similar feedback on the DSL side, about high-speed DSL flavors coming soon--ADSL2 and VDSL, to name two--would be welcome. I've opened up comments on this post.