Since there's no urgency here, it will be fine if it takes another generation or so to fully transition.

There are substantial costs involved in having to manage with two incompatible measurement systems. Never mind needing to convert between each one, how about every mechanic in the US needing to have two sets of tools for ever size, metric and SAE, doubling the cost. That's a huge, huge cost across the economy.

Since the rest of the industrialized world uses metric, the sooner the US completely switches over, the sooner we'll start seeing big savings from not having to be "multi-lingual".

And let's not mention the costs of spacecraft crashing into the faces of other planets.

No, a class action is the only way to go.

Except for the "no class action" clause in the AT&T's contract.

Well, we're screwed, then.

Those "no class action" clauses are likely unenforceable unless they give you a reasonable way to opt-out (and you're just too lazy to do so, like most customers).

Now let's just hope that the no-contract month-to-month plans don't disappear.

They won't... There's too many people on them, who would be willing to jump to the next provider at a moment's notice. If one of the big 4 dropped their contract plans, some tiny company would go around putting up cell towers, and trump them all.

Hell, WiFi tech is almost good enough that you could use APs to provide VoIP across whole metropolitan areas, for far, far less than the price of cell tower infrastructure. If competition doesn't get them, technology eventually will...

I pay like $78 per month for what on ATT or VZW now costs about $120 or so - I get a new phone and sign a new contract every two years

If that's for one phone, that doesn't sound like a good deal, at all.

Boost Mobile (Sprint) has unlimited voice/sms/data on Android at $55/mo, and has "shrinkage" which drops that by $5 every 6 months (if you pay on-time). After staying for just 18 months, you're only paying $40/mo, indefinitely, with no contract and no fees. And that's the TOTAL price... No surcharges, no taxes, no fees, etc.

As for phones, Boost has some decent ones for $80. Get a new one every 2 years, and you're adding just $3/mo on to your bill, for a grand total of $44/mo., or a bit more than half what you're paying for your wonderful grandfathered-in plan.

I wouldn't be caught dead signing a cell phone contract. If Boost/Sprint ever pissed me off, I'd switch to some other service immediately... No fees, no nothing. That's probably why they just DON'T DO THAT to their customers, ever.

I was under the impression that the issue of translating LED light into a broad swath of color was an already solved problem (except for some fine-tuning optimization), using appropriately-sized nanoparticles which hand the energy from the photons around, slicing-and-recombining energy from photons into different sized packets and re-emitting the light at a frequency characteristic of the size of the nanoparticle. Cover the LED with a bunch of these in a range of sizes and you get a smooth spectrum.

Works the other way, too: Coat a solar cell with such particles and they take the random-frequency photons from the sun and slice them up into multiple new photons at a frequency good for the solar cell bandgap, and mash the levtovers into more big photons to re-slice to the correct size. (It's not 100%, since some of the photons get away. But it's more than a 2x improvement over a bare cell, which only takes one slice off each photon and throws the rest away.)

If this is correct, this project looks like just a fine-tuning of making the nanoparticles, or finding materials for them that are somewhat more efficient than what was already being used (which was pretty good).

I haven't been following this all THAT closely. Have I misunderstood the current stuff? Or is this just a little incremental tweak along the cutting edge?

You use KiB instead of KB, don't you?

KiBbles & bits? Absolutely!

In some ways, the great success of Ethernet is that it became the name we gave to whatever technology won out.

No, ethernet remained relevant because it was able to improve, while maintaining backwards compatibility along the way, so your investment was never wasted.

10Base-T cards still had BNC connectors on them, letting you transition smoothly from one to the other.

100Base-Tx was backwards compatible with 10Mbps hubs & NICs.

Gigabit offered backward compatibility with 100Base-Tx.

Switching between fiber and copper is just a matter of swapping the GBIC/SFP transceivers in a switch, with the underlying device having no clue that the media is different.

Newer standards retained backward compatibility with older, less robust cabling... From CAT-3 to CAT-5, to CAT-5e/6, to CAT-6a.

Even though ethernet of today doesn't look like it did, originally. The upgrade path was always simple, smooth, and inexpensive, so it is very much an unbroken chain back to the beginning, and hooking up a modern PC to one of the first ethernet devices is a simple matter of physical-layer conversion.

Mother Theresa would no doubt have printed a medical tool for removing IUDs.

Which would have been totally useless since most of the countries and places she setup shop didn't have access to birth control to begin with.

India, with its huge population, had a large program making IUDs available at no cost to people in the poorer regions who wanted them.

Mother Theresa's work included providing medical treatment to the poor in many of these same regions. Her clinics were noted for removing the government-provided IUDs of women who were there for other procedures, without seeking permission or even informing the woman that it had been done.

Don't you just pull on the string?

Not if you don't want to kill or sterilize the woman.

CONFIRMED:

https://en.wikipedia.org/wiki/Vehicle_Assembly_Building

I've never seen "licensed" grid tie systems that didn't do what you describe.

And you won't: They can't be "licenced" if they don't do this.

About the only way you can feed the grid legitimately without such a device is by pushing on an induction motor (as happens sometimes in normal applications, like with a mo-gen system for an electric elevator when the elevator is being slowed down.) Induction motors depend on the grid for excitation and won't self-generate unless you've got enough capacitors hung across them (and your load) to make the net power factor capacitive. (This is unlikely but can happen in islanding, so you generally aren't allowed to hang a prime mover, like a water wheel or windmill blade, on an induction motor, and hook it to the grid, without adding such controllers. Generation from induction motors in an outage should only last for a few seconds while they suck the inertia out of some spinning mass and run down.)

Last I looked it cost about $2,000 extra to have a grid tie inverter with sell-back (and high-current load direct-connect with inverter "helping"), compared to an equivalent inverter from the same manufacturer that only fed the load with inverter output but could charge the batteries / feed the inverter with rectified line power when the grid was up.

In this case the $2k-ish bought you an extra box containing:
  - a contactor (to jumper the line to the inverter output) and
  - a circuit board that controlled the contactor and acted as a peripheral to the brain of the inverter, providing it with phase measurements (to line-up the inverter phase with the line phase before closing the contactor) and a voltage and frequency measurements (to tell the brain when the grid was failing, so it could open the contactor and cut you loose).

Though this was an add-on box, other products with the function built in also brought a similar premium compared to non-sell equivalents from the same manufacturer.

Straight grid-tie devices feed the harvested power to the grid and depend on it for interconnect and timing reference. Yes they don't feed a dead grid - but that means they don't feed YOU when the grid is down, either.

I'm maybe three years out-of-date on this information so the market may have changed.

Mother Theresa would no doubt have printed a medical tool for removing IUDs.

I do like the idea of a nitrogen atmosphere.

We're 3/4ths of the way there!!!

I wonder what kind of atmosphere conditions you could use to accommodate better cooling? Would a vacuum work better, or would high pressure work better for removing heat from the systems? Are nitrogen, CO2, Oxygen, or other gases better at transferring heat?

A vacuum would mean absolute NO cooling. The denser the gas, the more heat it could haul away, so something like argon would probably be best.

However, you could do much better by submerging the whole thing in Fluorinert or other (cheaper) non-conductive and non-corrosive fluid. The downside to that, being both that traditional hard drives will cease to function, and the weight of a building full of fluid will be astronomical, and would also require extremely tighter tolerances and far more horizontal support.

The only way I could see that working, would be a huge subterranean datacenter... Basically a huge hole in the ground, or perhaps the world's deepest in-ground pool.

I've been in plenty of datacenters, and I don't see where you're getting any benefit with radical redesigns. They aren't exactly designed for human comfort in the first place...

Lighting? Sure, but motion sensors mean it's only on when someone is in that area. And you'll still need lights, because humans will surely still be going in there to fix the malfunctioning robots, and hiring old coal miners seems excessive.

Temperature? No, the servers dictate the temperature the datacenter is kept at, while human comfort is completely secondary. The 15C degree air coming out of the floor vents below my KVM doesn't make for a comfortable experience, but nobody cares. Humans in the datacenter are the foreigners, who must adapt themselves, not the other way around. If Google could run their datacenters at 75C degrees, they WOULD do that now, and the humans would be sent in with ice packs strapped to their bodies.

Height? If a couple more feet of rack height were useful and cheap, I would be happy enough to keep a bit of scaffolding in my datacenter cages. As for the ridiculous heights predicted, it's not going to happen. Racks can't scale-up that easily (they'd need huge thick vertical supports to handle the weight)... and at some point, it's pretty easy to just install another "floor" for those pesky humans to walk on, install air ducts in, and also avoid the need for super-robust racks... and I can't even imagine that crazy air currents that would be happening with 100' of vertical servers pumping out crazy amounts of heat, not to mention problems like CLOUDS forming and potentially raining, INSIDE the building.

In general, the comparison needs to be made to warehouses... If Amazon/Walmart/etc. had fully-automated warehouses, I'd say automated datacenters would be just around the corner. But they don't... Humans are still very much in the loop, driving around on electrified forklifts or pallet jacks, and doing what the computer tells them to, and when. And if any business could benefit from vertical expansion, quicker response times, and less humans, it's warehousing, but it just doesn't work there, yet. That will be a lot closer to the model for future datacenters, not this pie-in-the-sky nonsense.

Even if only a third of the people stick around after din-din (and it's usually more), it's still the equivalent of getting more better than a 10% increase in manpower for the price of nine dinners (in bulk) per day per extra head - FAR less than the cost of hiring another head.

Did the math wrong: Make that about 17% more "heads" for the price of six dinners per night for one in three staying an extra half-shift..

FWIW, most "free food" programs encourage workers to come in earlier (for breakfast) or stay later (work past dinner time) or to not spend a long time off the company property over lunch. The extra time at work usually pays for the food costs.

It pays MUCH more than that. A typical thing that happens at startups is the company buys dinner - and the bulk of the engineers chow down and stick around another four hours. Not only do they get half-again as much time, but they get it in a block. For a programmer or other design engineer that means they haven't "lost state" and are even more productive than if they'd just worked three days instead of two.

Even if only a third of the people stick around after din-din (and it's usually more), it's still the equivalent of getting more better than a 10% increase in manpower for the price of nine dinners (in bulk) per day per extra head - FAR less than the cost of hiring another head.

And then there's an adminstrative pathology: The new management comes in, sees how much is spent on the food (but not how much is gained as a result), decides that their predecessors were stupid and the employees were looting the company, and stops the food. So come dinner time the employees go out (or home) to dinner and don't come back. Immediately it's like they lost somewhere between 10% and 33% of their work force without any reduction in payroll costs. (That's not counting how disgruntled some of the employees become.)

I've been at three companies where this happened, and observed several more. All but one of 'em folded shortly thereafter - and the one that survived went through a near-bankruptcy that destroyed the original investors' equity and left it in the hands of the bondholders before it recovered.