Last week's column on the death of broadband brought a stiff and generally negative reaction from readers, though there was a small contingent of cynics who agreed with me, and even provided further details why broadband, as we know it today, isn't especially viable. Most of those in the pro-broadband crowd were happy cable modem or DSL users who didn't want to lose their cheap connections. They tended to see broadband as viable because they wanted it to be viable. They reminded me of the scene in "Peter Pan" where we all clap our hands to bring Tinkerbell back to life.

This week, we'll revisit this topic, and I'll try to both clarify some points that were apparently muddy in last week's column, and bring in some new points made by readers on both sides of the issue.

Let's start with a question. Why does a T-1 connection to the Internet cost so much more than a DSL connection of similar bandwidth? Who would buy a $1,500 T-1 when they could pay $300 for a 1.1 megabit-per-second symmetrical DSL line from Covad? The T-1, at 1.544 megabits-per-second, is slightly faster, sure, but the cost is five times as much. Why are there any T-1s sold at all? The answer has to do with both proximity and reality. In order to qualify for that 1.1 megabit DSL line, you have to be quite near to the telephone company central office. DSL bandwidth drops off with distance, while T-1 bandwidth does not. So even though it costs more, T-1s are viable high-speed connections for far more locations than DSL will ever be. But at five times the cost, it still seems absurd to buy a T-1 — that is, unless you actually want the bandwidth you are paying for.

With a high bandwidth DSL line, you are typically paying for speed, not throughput. Run your DSL connection flat-out for a couple weeks, and your ISP will put a throttle on your connection or try to charge you more. That's because, as I keep saying, they can't make money at the current prices. DSL and cable modem providers still have to buy bandwidth to the Internet backbone at the same price — a price that isn't typically justified by the low prices they are charging end-users. They are counting on customers who are occasional or sporadic users, and who primarily download, rather than upload. That upload limitation is to avoid that most hated of all things to a DSL or cable provider — the user who runs his or her own web server. An even modestly popular web server can eat bandwidth equivalent to hundreds of regular DSL or cable surfers.

Download bandwidth is less sensitive for two reasons. First, many broadband ISPs run proxy servers to cache frequently accessed content and conserve bandwidth to the Internet, itself. Second, there is the throttling effect of overworked servers. If your favorite web server at naughtynurses.com is set to accept 500 simultaneous connections and its connection to the Internet is one megabit-per-second, it is very hard to connect at greater than 20 kilobits-per-second no matter how fast your DSL or cable modem is. I first noticed this effect when I increased my DSL connection from 384 kilobits-per-second to 768 kilobits-per-second, and saw no effective increase in speed except for some (not all) FTP downloads.

This all has to do with the economics of the connection between your ISP and the Internet without regard to what kind of connection you have. It's just that only the really fast connections are in positions to notice the effect. And it is going to get worse, not better, because bandwidth, contrary to popular belief, is getting more expensive, not less.

How can this be? Isn't the Internet based on the idea the total bandwidth doubles every few months and the cost per bit is ever dropping? That is certainly the popular idea and it has some theoretical support, but reality — the new reality — is somewhat different. There is an Internet food chain and just as you have been buying below-cost bandwidth from your ISP, your ISP has been buying below-cost bandwidth from its provider. During the Internet bubble, especially, Tier 1 backbone providers cut deals with ISPs that either sold bandwidth below costs or, more typically, deferred any payments at all for a year or more. Rather than a one-year contract, ISPs often took three-year contracts in which the first year was free. That's a heck of a deal, and one that made sense at a time when none of these companies had to be profitable to be loved by Wall Street. Well, the bubble is gone and so are many of those ISPs, some of them departing without ever having paid a cent to their bandwidth providers. The bandwidth providers are still here, for the most part, but now they have to absorb the losses for all those unpaid-for bits and bytes, and simultaneously deal with the new reality that Wall Street now demands profits. So bandwidth providers are effectively raising prices by not lowering them. Bandwidth costs have stabilized in what ought normally to be a declining market.

Now to the readers, none of whom wanted to give up their fast connections. And as long as these users are content to accept speed rather than throughput — as long as they don't expect to download movies — they will probably keep their cheap connections. The result is a class of surfers who don't do anything especially different from the rest of us, they just do it quicker. Good for them.

I heard from a number of successful DSL ISPs, too, though every one of them was a BUSINESS DSL provider. None were claiming to make money on $50 residential connections. I heard from a number of wireless ISPs, too, though those are just as subject to the backbone connection cost issues and have scalability problems, too. MMDS (Sprint Broadband) may be great, but it is inherently limited to about 15,000 users per cell with each cell covering abut 700 square miles. Those numbers may work fine in small towns, but in major markets, they doom MMDS to being a minor player.

Some readers lamented the loss of higher bandwidth content. Here is a rant so heartfelt I have to quote it directly: " Since I use the net for all my media (I don't have television) I notice when things change. I have noticed the following changes since January relating to news. Foxnews has stopped all Internet feeds with only a big advertisement for a web site. CNN has stopped all live feeds and offers only canned segments, quite boring and time consuming to load and watch. MSNBC has also stopped all live feeds except for important events, and they have limited all their feeds to 15 minutes before it just stops no matter how far along it is. ABC and CBS did not have much of a live feed system in place but have been adding to the 'canned news segments' format. Premiere Radio has eliminated Art Bell from free feeds while paying Limbaugh some incredible amount of money for his contract (somebody has to pay for that, folks). In addition to that, nobody is expanding and improving on Internet services, and everybody is cutting back. It is a death wish in real time. Live feeds are almost non-existent."

In a network with millions of channels, can it be that there isn't anything on worth watching?

Another reader asked, "Wasn't it exactly a year ago that you were touting "broadband" as the "savior" of the PC business?"

What I wrote back then was that it would be in the interest of the PC business to do whatever it took to make broadband succeed. Just throwing a couple billion per year into the business to cover those losses would have kept growing a $200 billion PC industry. But hey, did anyone listen to me? No.

Finally, there was one reader who gave some hard reasons why telephone companies might see DSL as a profitable business, though indirectly. This is really interesting: "You missed something when you said that ILECs aren't making money off broadband DSL. This is simply that by deploying DSL the ILECs are getting the worst modem abusers off their phone lines. These are the folks that leave modems dialed into ISP's 24x7 or more than 200 hours a month. When the ILECs (convert those users to DSL), it reduces the need for them to build voice switching infrastructure. In short, when they get a city converted over to DSL it reduces the average voice call time. So while they may be breaking even on DSL or possibly losing some money on it, that is far more (than) made up by the enormous amount of money saved by not having to increase voice switching infrastructure. A DSLAM is a lot cheaper than a new phone switch. One other thing you forgot, too, or don't know — in many areas the T-1 delivery capacity from the CO is maxed-out. Well, from a phone companies' perspective, it is much more profitable to have 200 T-1s all with voice service on them sold to 200 customers, rather than 100 T1s with voice and 100 T1s carrying point-to-point data or Frame Relay. When DSL is deployed it reduces some of those T1 circuits, freeing them up to be used to deliver PRIs, which are much more profitable."

In the end, the phone company always wins.