I have thousands of photos "in" iPhoto, and I have no idea of any of the file names or their location on the file system.
I don't care what they are called or where they are, I just want to have my photos.
Sure, there might be a filesystem under there, but I have no interest in interacting with it directly.
Normal users seem to be very comfortable with the idea of unspecified magic though.
To try and paint computer users as being in two camps 'normal' and 'developer' does everyone a disservice - there's much more variety in users than that. Here's an example - there's a lot of gamers that like side-editing save games, yet they're not developers. Where to they fit into your dichotomous view?
I think 'regular' users have plenty of use for understanding the filesystem. I also think that it's okay to have some degree of expectation for the user - dumbing it down for the lowest common denominator is harmful (look at politics!). What do you really gain by alienating the power users?
I see no reason why making the common case simple (not "dumb") means you have to eliminate, or even alienate, the power users. Frankly, since I started using a Mac, iTunes and an iPhone, I stopped managing music files. And I think that's great. Files are an implementation detail. What I want is music or radio programs, not files.
The idea that the 'normal' user is only a home-based casual web-browser/photo-taker/music listener is misleading from the outset.
Even a flickin' unix whiz doesn't care about, say, which disk sector a file is located on. In fact, if they are using a distributed file system, they also can mostly ignore the physical location of the data.
Your spiffy net of tags is a way-to-find data. A officially "file system" is a way to find data. Each has some flexibility, some lack-of-ambiguity and some persistence over time. You may not care about the uniqueness, ambiguity and decay over time of your tag based way-to-find data. But that just means you've got a convenient but half-assed file system. Moreover,
If tags are a way that you can definitely find and specify a given file, then tags will form part of a new, distributed file system. If tags wind-up just being half-assed, uncertain hints to where files might be, then they will form part of a new, disfunctional distributed file system. And the later case seems to be where things are going. This "hints but no certainty" approach to file location indeed works great most of the time and fails frustratingly significant percent of the time.
But even here, this is a file system even when it is often dysfunctional.