The trouble with them was the cost of the mercury, particularly during war time.
Turing once suggested that "Mercury Line Memory" could be made with Gin... --I've always wanted to try it.
There is some analogy here with bubble memory (though that moved around physical material rather than just set up a wave).
I wonder what the minimum spacing of such waves would be, how this would compare with current estimates on the upper bounds of density for silicon and if material fatigue factors into this somehow. I can't imagine such a system to be very stable mechanically.
I'm really not sure, and I'd rather not speculate blindly. I've read
about the basics of Mercury Line Memory in the past, and have some
books that describe it, but I don't know the hard details.
BTW, Mercury Line Memory was not the first form of memory used in
programmable computers. The "oldest" or "first" is debatable, but
the earliest were switches/relays and vacuum tubes (AFAIK). I'd
bet jgrahamc would know...
I don't think it was standing waves. If memory serves (i.e. I may well have got this wrong) they stored data as a pattern of pulses travelling down a mercury tube.
They borrowed this trick from radar engineers. The engineers needed a way to compare what the radar is seeing now to what the radar saw on its last rotation, so that they could detect moving objects. They did this by sending the information down a mercury delay tube of suitable length.
You're right, it's not standing waves, being used a delay line. You make it 'persist' by sending a signal down the delay line, reading it out the other end, and regenerating the signal into the delay line. This is different from a standing wave, since a standing wave results from the reflection of a wave from the ends of its transmission mediums, while in delay line memory, reflections would probably be a really bad thing.
Relatedly, if you ever had an old school analog sound mixer with sound effects, it would have a spring delay line for delay and echo effects. If you were clumsy like me and ever dropped one, you could here it boinging around.
So given that developers have a lot of trouble programming normal simple binary computers even using languages as high-level as python (not to mention their trouble at grasping advanced languages like Haskell), why do we think they will do better with fuzzy logic for which we don't even have high-level languages?
Found it. http://en.wikipedia.org/wiki/Parametron
This also might be of interest http://spectrum.ieee.org/computing/hardware/dudley-bucks-for... http://en.wikipedia.org/wiki/Cryotron
For completeness there was a great talk the UW EE department a couple days ago, https://www.youtube.com/watch?v=cwl6fORHNqs on using electron spin to store information.
https://www.ee.washington.edu/cgi-bin/research/colloquium/di...