I can work with that, though it is worthy of note that programming as a "use" or "interface" is a different (in a significant sense) beast from some other uses (and I will argue, this difference, a sort of Russellian Theory of Types thing, cannot be entirely removed).
> We can immediately see that there are better and worse programs. Some
> let us think the way we want and offer helpful features; while others
> impose distractions which have little to do with what we're trying to
> describe. We prefer to describe games and accounting systems, not
> what's supposed to go in memory location X.
Which is agreeable to, but again with a caveat: each of us [likes/ wants to] thinks in different ways! And there are always trade-offs related to complexity, the issue that the IP over XML RFC parodies.
> Why aren't we using machine language and Cobol, if we can help it?
> Because languages greatly differ in expressive power. The thoughts you
> can conveniently express without fighting the language.
The reverse of that question is also legitimate: why are we using C or Perl or Assembly? Because, after many an attempt to do without, we find that we cannot! I believe there is a bit of truth to both sides: languages can be a lot more programmer/user-friendly but also complex tasks require complex expression. The latter was part of the point I was trying to make about real world code.
I have trimmed out your bullet list of attractive features in a programming "interface" (I assume that when you mention garbage collection in the last point, you meant that the computer would do that), since I tend to agree with them.
> And Alan Kay, developer of Smalltalk and (I believe) heavily involved
> with One Laptop Per Child, claimed that "the computer revolution
> hasn't happened yet."
I think what is missing in the above sentence is: "... and it never will". For example: lay people tend to think of the Internet (or World Wide Web, as it is called in the media ;-)) as a revolution, but it is quite evolutionary and a slow evolution at that... I remember Kernighan or Tom London telling me that networking as exists today over IP was envisaged as a natural consequence (Datakit notwithstanding) back in the early 70s. I do think there will be lots of "revolutions" of this type; things that can exploit division of labour (through clear boundaries and so on), but not things that bump up aainst the inherent complexity of problems. And so we have as a result of the 90s not The Fifth Generation Project (though we have a lot of good heuristic engines and chess programs) but TCP/IP. IMO.
--ravi