I would have thought that the fact that the experiments with leaves brought there from elsewhere decaying slower demonstrate that merely bringing foreign organisms (the collected leaves are not sterile, of course) is not going to help.

It's a full set of organisms you need; if, for instance, the earthworms were missing, a few strips of sod (or waiting for 'foreign' worms to migrate in) would be effective, but a pile of leaves wouldn't.

While there may be millions of possible reconstructions for a fuzzy, ill-defined image, the simplest (sparsest) version is probably the best fit."

Of the millions of possibilities, the sparsest is MOST likely. Perhaps it's twice as likely as any other possibility. That still means it's 99.999% likely to be wrong

I interpreted this to be a description of maximum-entropy filtering (i.e. making an output image with least information, consistent with input image with sparse information content overlaid with full-textured noise.)
It certainly works for (for example) starfields seen in a badly focused telescope; NASA used this for Hubble reconstructions.
Every possible reconstruction will have some 'wrong' content, because noise is guaranteed (if only by the quantum uncertainty principle). That doesn't make the maximum-entropy filter useless, it just means the utility depends on the subject's compliance with the minimum-information assumption.

Frozen ground is only waterproof if there are no holes. Frost heaves tend to break up the ground and make holes.

Actually, a frozen region in the soil is a good container, because water that starts to
  leak through... freezes solid and plugs the leak.

Frost heave is caused by thermal gradient, and
transports water to the coldest spot (which is
the container wall, safely underground) then freezes it.
So, no problem there!

Any cast metal needs machining afterwards, even if it's just to clean up the sprue area and to cut screw threads. Difficulties in casting are not why we don't see conventional metals used ...

Yes, and no: most cast metal shrinks markedly when it cools, so there's an oversize pattern (two percent for iron) used in making the mold. Amorphous metal has almost the same density when it's cool, it shrinks only slightly (an order of magnitude less shrinkage).

credit for the invention belongs to Dr. Joseph V. Foa who was awarded US Patent 3213802 for a "train in a tube" in 1965. This was the basis for a number of years of research into the concept at Rensselaer Polytechnic Institute in the 1960s.

It's far older than that, of course. Isambard Bunuel was tinkering with 'atmospheric railway' hardware a century and a half ago. Patents issued in Britain, 1838.

A modern day Ben Franklin would outsource the kite flying: some inexpensive foreigner takes all the lightning risk and Benny gets all the credit and fame and votes himself a raise.

Odd you should mention this. In fact, Ben Franklin described the make-a-spark-in-a-storm experiment, and some Frenchmen hired a retiree to do the work. Ben was presented to the (then) King of France with the results (and suddenly, America had a bit of international prestige, no longer just a backward colony).

[it simply doesn't cost THAT much]

It depends on what they are including in that cost and how they are amortizing it. For instance setting up a local relay station for a small town...

But the cost per CUSTOMER in that small town wouldn't ever be so high.
There's microcell pole-top repeaters that can run on a solar panel and battery, that can handle the 'last-mile' problem in a rural environmen. Cisco 1000 series, for instance. The lesser monetary amount mentioned ($3k) would buy one and its maintenance for ten years. The last-mile problem is solved at the community level.
Low-capacity fiber backbone is relatively easy to build, and should be provided for in any state roadway construction. That solves the 'last 40 mile' problem, at the state level.
So, the real issue is 'last-500-mile' connection, and that's a necessity in interstate commerce, which we've presumably already solved independent of the subsidy.

I'd disbelieve any '$9k per year per customer' costs, and that means either the researcher is cooking his numbers, or there's a telco fraud (like, first-year retirement of costs on a 10-year bit of infrastructure).

30 years ago I was using a 3 1/2" floppy drive. I still have one to this day. Plugs right into a USB port. In fact, you can still get hardware to read any of the old media that you stated.

Not true, of course. The Macintosh floppies of 1984 were recorded in zoned fashion (required a variable speed disk drive), not compatible with 1.44M modern 3.5" drives, and the early file formats aren't intelligible to modern computers. There were so many 5.25" floppy formats that we had a special computer set up just to read 'em all (does anyone remember Discon?).

Thankfully, there's a good written standard for CD and DVD filesystems; those will be readable for a long time if the media survives.

No need to worry about ink: even the cheapest and nastiest laser printers use toner, and a mixture of thermoplastic and carbon black thermally fused to your paper isn't going anywhere...

Carbon black does go away (turns to carbon monoxide) over a few centuries; the run-of-the-mill toner, though, uses Fe3O4/Fe2O3 pigment (it's slightly magnetic). So does classic oak-gall ink for quill pens, and the longevity is good.

Not the issue; this is about keeping a backup of some
currently-in-use data, it'll get REWRITTEN onto those
flash drives every year or so.

So, a ten year data retention is more than adequate.

It'd be really easy to come up with a suitable fire-safe
place for a dozen microSD cards; that's only five
cubic centimeters...

Mac Air:
2xUSB 3.0, HP/Mic, SD (Air 13), Thunderbolt

Ativ:
2xUSB 3.0, micro HDMI, mini VGA, RJ45(Dongle), SD, HP/Mic

Ativ beats air by 2 video outputs and wired ethernet. Also by SD when compared to the Air 11.

The Thunderbolt contains miniDisplayPort i.e. DVI and HDMI and VGA all in one;
just choose a short adapter or use long cable with two different end-connectors.
Maybe you can call the tiny connector 'nonstandard', but it supports two distinct
standards (DVI and VGA) as well as some variants like HDMI.

The 'micro HDMI' isn't any more standard, in any useful sense. The real difference,
  is only the missing wired Ethernet port and maybe SD card slot. Neither gets lots of use on laptops.

It's not practical or safe to fuel a car with pressurized gas or with liquid H2. The best prospect for automotive hydrogen tanks is a sponge-like intercalation storage, and it TAKES TIME to fill (or drain) that kind of tank.

H2 cars will have to swap cartridges at refuel stations. Metal embrittlement is no real problem when the fueling stations are required to reinspect and rebuild on safe schedules.

And WTF is proprietary technology?

Technology that belongs to Apple and is incompatible with everything else. Many other machines used standard floppy disk controller chips.

That's twisted; proprietary technology means OWNED technology. Apple had a patent on some its floppy controllers, and IBM decided on a NEC part, uPD765 if memory serves, that was proprietary to NEC. It wasn't a standard, either, just a documented solution that DOS was made compatible with.

All the early floppy disks were proprietary. CDs had a data-format standard, though.

A standard is built around a full formal specification, by a group (IEEE, ANSI...) that usually is not the 'owner' of the underlying patents, but is an interest group which sponsors the publication and growth/modification of the specification.

>I suspect that you are likely going to be very safe with Cat-6 for a few decades, but ...

A few decades is the human lifespan. That's good enough for me.

Parenthetically, cat5 was developed for the old only-uses-two-pair standard, 10baseT, but works fine with four-pair gigabit (1000baseT) gear. The cat6 advantage isn't really speed, it's range (and for my house, one can painlessly forego any range beyond 50m).

>Have fun writing...algorithms for those if you can't work out analytical solutions

Truth. It's easy to write code, but only a good little
computable-on-back-of-an-envelope test case will give
me confidence in it. I've found errors that were hidden
four decimals down in my brute-force algorithms.