Thanks for the detailed response.
> Automatic landing systems rely on ground support by way of microwave signals from transmitters installed near autoland-capable runways. There's no general position information, and the airborne part of the system can't operate independently of the
part on the ground.
Yes, I understand the landing portion has other kinds of support. However, I don't understand how it can not have the positioning information we need, because {see next response below}.
> Civilian GPS signals are limited in terms of position accuracy, too. I believe the CEP is ten meters. That's not bad, but it's not good enough to land an aircraft. Altitude information from GPS is (IIRC) either nonexistent or severely limited, too. About any aircraft with a glass cockpit is going to support a moving map display, and those do see use - but, as a couple of professional pilots in this thread have noted, during final approach the main concern with regard to aircraft position is relative to the glide slope and the runway, and a moving map has nothing to do with any of that.
This seems to contradict with what I read [1]. See this quote, for example:
Generally, the pilot will handle takeoff and then initiate the autopilot to take over for most of the flight. In some newer aircraft models, autopilot systems will even land the plane.
OK, so I don't have much of an expectation for older plane models, but my interpretation of the above is that in newer models, autopilot need not disengage before landing and require manual intervention in order to align the plane with the correct track (barring bad weather or other unusual situations like bird collisions, etc.). This is supported by the fact that they later explicitly mention that autopilot occasionally disengages. So it seems clear in normal situations autopilot has plenty of positioning available to it, enough to find the runway and locate the plane on a display of some sort. That goes pretty clearly against what you said. Am I missing something?
> On a more general note, it is remarkable to me that so many otherwise apparently sensible software engineers seem to regard aviation as a problem trivially susceptible of perfect solution
I hope this isn't referring to me, because I've been trying pretty hard not to give that impression (which I don't have). I never suggested anything about aviation perfection in general. I'm specifically talking about this particular kind of problem, because it seems to me that with current technology, pilot errors of this kind should be completely avoidable using current technology and, basically, non-problems in the first place. (Again: note that I said these types of pilot errors should be avoidable. I didn't say all faulty landings on the wrong tracks should be unavoidable.)
In other words, nowhere did I suggest that I think airplanes can fly or land themselves perfectly in unfavorable conditions. But this was quite a favorable condition for an automated system.
And, as a computer scientist, what seems ridiculous to me is the idea that "Oh, but this system you just described won't work if there is a hurricane and there are turkeys getting rammed through our engines while the pilots are asleep and ATC happens to be on strike in the middle of a nuclear holocaust, so let's "be conservative" and never approve it because it's clearly going to make things worse on average".
[1] http://www.cnbc.com/2015/03/26/autopilot-what-the-system-can...