Digital goods do seem to be doing some wonky things that our previous models don't handle well. There is a combination of near zero marginal costs, near instantaneous global distribution, and near zero transaction costs. Before the internet, books and movies did not have those three properties.
In a way, you could argue that iTunes was the fracking of music sales, and that other markets need something similar.
Instead, the startup costs can be paid for in a collaborative framework (open source, wikipedia, and so on) and the growing snowball of results freely used and remixed in common by all, producing more results.
Then the marginal costs would be simply hosting and support. The questions of “but then who would make XYZ” have already been answered.
The question, though, is will we still have famous authors and heroes, or will our digital products be made 95% by the crowd. I think the latter. Wikipedia and open source have very few famous personalities - just the BDFLs that launched and run the project, and not specific creators of content.
With the printing press, this cost went down, but it still took a significant effort to create a copy; giving that copy some independent value.
Digital media has practically zero marginal cost. So long as you are connected to someone with data, it is trivial, and practically free to copy that data.
There are efforts (DRM) to fill the void of marginal cost; but they are hacked together, and fundamentally flawed.