Thanks. And I substantially agree with you, but it's more fun to disagree, so I will.
The only User Interface guru is your mum.
There are three stages that people go through when they are getting good at this. At the first stage, they think they know everything. At the second stage, they think their customers and stakeholders know everything. At the third stage, they realize that it's very important to listen to the customers and stakeholders but also to know when not to do what they ask. Expecing your mum (or mom for the U.S.) to be a User Interface guru is like expecting a film critic to write or direct a film. If you don't see how that isn't necessarily a good idea, see if you can find Myra Breckenridge by Roger Ebert.
Now have you noticed the subtle flaw? What if you wanted to work on more than one document?
It needs to be noted that the biggest proponent of getting rid of save is Alan Cooper. As the "father of Visual Basic" (that's what it says on the blurbs of his books), he is probably indirectly responsible for more bad user interfaces than anyone else on the planet.
That having been said, the problem with Save is not that it is bad to have a time when you could finalize your changes, experiment with things you don't intend to save, etc. but that it was not designed to do that. It was designed to solve the problem of RAM versus permanent storage. Nobody sat down and thought about the actual user problem and came up with a good system and good names. But when Save came with all programs, people hammered the square peg into the round hole and used it for that reason.
The trouble is that because nobody has really thought about the problem, the alternatives suck. "Make everything undoable" is the usual response. Well, have you tried to remember fifteen levels of undo? Undo is just fine for changes you've just made, but it sucks for anything else. Besides, it doesn't deal with branching. I've often thought that a kind of date-tracking mechanism would work (i.e. show me this document as it looked on June Thirty-Seventh in the year of the Squirrel), but nobody has sat down and worked through the system.
HyperCard, which was enormously successful in spite of the fact that hardly anybody remembers what it was, used a saveless paradigm, where one could save snapshots and then later go back to a snapshot. The weakness of this is that you seldom know that you're going to have to go back to a snapshot before you mess up. Also, a game called System's Twilight (an excellent puzzle-solving game, BTW) was saveless, but then again it had to be or the game wouldn't work.
Filenames are so passe...
...now let me just enter ten key-value pairs in a database.
False dichotomy, really. But again, file names weren't designed to solve the problem of finding files in a huge system. They were designed as unique keys for relatively wimpy file systems, which were extended for technical reasons to hierarchies. There are a lot of ways of doing things better than filenames or relational databases. The Be OS metadata concept was very good in this regard. There's even one simply change that could be made to existing file systems that I think would be better than the status quo: allow a file to reside in more than one folder. Yes, you can kludge this with aliases and links, but that's the save problem all over again.
Well, because whenever you looked for something, you'd have to trek up and down into all these subfolders, remembering how your mind worked 6 months ago.
Hard as that it, it's not as hard as predicting what your mind is going to work like 6 months in the future, which is what the filenames ueber alles approach requires you to do. I'd like to be able to find files that I don't remember much about, but I remember it was before last year's Siggraph, and I worked on the Flebeski account around the same time. I could do this, but I don't because no file finder makes it easy. This will take some work, but it isn't hopeless.
It's time for a total rethink! Er... er... virtual reality 3D walk through file managers! Or a CLI, but like a new CLI, that's new.
It may surprise you that the uncontested Uebermensch of UI gurus, the great Donald Norman himself, considers 3D file managers ridiculous. Or not, as you acknowledge that you're setting fire to straw men.
There are plenty of good rethinks that have been done and could be done. I recently bought a Wacom tablet. The stylus has an eraser on the end. Great stuff. And there are certainly good pieces of research out there. The Teddy interface and SketchPad seem to have hit the sweet spot for some kinds of 3-D gesture interfaces.
The big problem with the WIMP interface, seldom mentioned, is that it's really good when you have small screens and simple programs. However, with several hundred menu items and dozens of windows and millions of pixels, all of a sudden the basic idea of having everything visible at once doesn't seem so hot. I'd like to see more interfaces that emphasize certain kinds of information, such as the zooming data path and cylindrical menus researched in the 1990's. As far as I can tell, the Mac's zooming doc is the only common use of this, and I like it a lot. (I mean zooming while running the pointer across, not the hateful Genie effect.)
Another bugbear of UI gurus, who want every OK button to say something different: which is like having a different type of handle on every door in your house.
Again, it may surprise you that Apple have always told by people never to change the name of the OK button. Or, again, maybe not. Apple, for all of their faults, spent an awful lot of money hiring people to work out a lot of stuff, and they did a pretty good job. They actually added value to the WIMP interface. However, one of the many things that Microsoft and the Free Software/Open Source people have in common is the idea that it's good enough to say, "Uh-huh. Look, Beavis. It's a widget. Cool."
For myself, I say that there are instances where it is appropriate to get away from OK, but they're rare, and they should be thought out very carefully. Often, rewording the text works better.
And WIMP is, I think, misplaced for mobile devices - one reason why I would back the mobile phone companies against the PDAs. (Have you seen the O2 XDA with Pocket Windows? You can right-click in it, FFS.)
Pocket Windows, Windows CE, and all their bastard brethren just basically suck. The Palm OS is much better. Of course, that's why Palm isn't doing so well these days.
The truth may be out there, but lies are inside your head.--Terry Pratchett