But, there's more:
The article emphasizes that the "costs" for an Internet Service, which hitherto have been cross-subsidized by advertising are, for all practical purposes, not relevant any more. It is true that this article talks only of the bandwidth, storage and computation costs and excludes the cost of labor. But much has been said before about the costs of labor decreasing via improved frameworks and tools.
The question then becomes, "What does it mean now, that the costs to be cross-subsidized are so low?" This goes back to what PG, Marc Andreessen and others have been claiming all along: we are going to see more startups, a LOT more startups. Also, according to Chris Anderson's Long Tail, LOTS of startups in niche domains. Startups operating in niches, might not make a whole lot of money through advertising. But, this is precisely the point of the article, given the cross-subsidization needs being so low, some startups would be happy making a couple of thousand or tens of thousands a month (instead of millions).
YC uses this as advertising.
Money isn't really wealth, money is just a representation of wealth. At some point, the overhead of using money will not be worthwhile.
A very hypothetical example: say you have millions of simultaneous users all over the world who are constantly, but unpredictably, using your web service. The value of all these people using your service together is very high. However, since people use the service irregularly, charging them a significant amount would make it not worthwhile for them to use your system, and you lose your crowd. On the other hand, the amount that you could reasonably charge them is so small that it would only cover the cost of transaction, and would make the psychological overhead great enough that you would again lose your crowd. So, overall, it is best to not charge anything, and instead extract value in other ways, such as demographic information mining.
Better to talk about currency than money. You are right, its place is where what you want to consume and what you (want to) produce are different goods. I don't see why the web would change that basic principle. You may contribute your expertise to Hacker News, an open source project and a couple of Wikipedia pages. But surely you don't want to consume only your own products. But maybe the producers of your desired "goods" (whether purely informational bits and bytes or real physical apples or oranges) derive no direct benefit from your aforementioned products. You will necessarily arrive at some medium, whether you call it credits or miles or dollars.
You also see this realization in the current trend of Services Oriented Architectures, which is basically the UNIX philosophy writ large.
Money has value because we can receive it in return for value we've created (a sort of universal value system), and transform it into another form of value or wealth. I don't see how it could ever get to the point where it's not worth having a universal medium of wealth. If it did get to that point, we'd have to resort to bartering. The only alternative is to have EVERYTHING be free - not just information products, but physical products, like cars, houses, and services. And if communist countries are any clue, this isn't the best way to do things.
The faster, more fine grained, and more dynamic a system is, the less money makes sense, since the cost of managing the money comes closer to the value of the money being managed.