They are not trained to protect your life. They are specifically trained to desensitize them to one's natural aversion to killing a human being. They are trained to protect their own lives by killing you. This is their reason for existence, to kill people

You have to train anyone who might have to use a firearm that way, because many people naturally won't immediately aim at a human target and pull the trigger even if their life depends on it. Most of the ones that will, will forget to aim properly, and miss... requiring more shots to defend yourself, and with every extra bullet fired risking killing some bystander.

Anyone who is considering owning a gun for home defense (or in states where you can legally carry for your own self-defense outside the home, generalized self defense) should take similar training. Get used to shooting the gun, then go to a properly equipped range with pop-up or flip-on targets which are photos of real people not just silhouettes, and practice shooting those too.

Officers, and anyone likely to be going in to and having to deal with ambiguous potentially lethal but usually not situations, should also get the video based shoot/no shoot training. Officers who have good basic firearms skills typically shoot a few perfectly innocent people, and get killed by a few innocent looking lethal people, in those training scenarios, before they internalize procedures to protect themselves.

Any police officer who does not come out of that training with an acceptably safe approach for dealing with situations on the street is likely to be let go by the department. Because if they will not shoot when their life is on the line, they are placing other officers' lives and innocent crime victims' lives at risk. If they will shoot inappropriately, they will wound or kill someone who didn't deserve it, and wreck themselves mentally, and get a huge lawsuit against the department.

A vast majority of the people police shoot are career criminals, who have guns or knives, and who pose an immediate and obvious risk to the officers or civilians. See for example the nutcase in New York City a couple of days ago, who was threatening and scamming tourists and when stopped by police pulled a MAC-10 (semiautomatic) and started shooting.

If police can't stop those people, then we're all screwed.

If we let the police turn into those people, then we're also screwed - but isolated incidents are not the totality of what police in the US are. Most are out to protect us, not screw with us.

Exactly - hypothermia in under-ice immersions, and surgical hypothermia, are well studied now and effective.

You can't possibly do that on the battlefield... but, with these chemical equivalents, turning off the oxygen for the decay processes rather than the temperature needed for them to occur, it's a possibly field expedient method of achieving the same goals.

Not magic - may not work if the blood's lost too fast or heart's not beating by the time you get to them - but CPR may be good enough circulation, and possibly IV bags with the chemical cocktail and saline solution would be enough to perfuse the patient even if most of the blood is gone (if you're putting them out, then no oxygen in the saline is not even vaguely a problem... it's actually helpful).

It may not end up working on human sized animals. That's what the trials are for. But if it does, it could save not just soldiers but many many many people who die in trauma or in surgical incidents.

Have to test in human sized animals, if that works then go for humans as with any other medical procedure. No way of knowing ahead of time. Scale effects may screw people sized patients. But we won't know until we try.

"Fogbank", widely presumed to be a heavy-metal doped aerogel material.

We can manufacture it again. There was a gap - we couldn't for a while, but it's back in production.

https://secure.wikimedia.org/wikipedia/en/wiki/FOGBANK

Good on you. You get it.

Academic programs often have an unfortunate tendency to turn out people educated like they were going on to be academics.

That said - Unawareness of the wider world of algorithms (and wider world of Computer Science, writ large) is a self imposed glass ceiling in the programming field.

The real key is not whether you went to school. It's whether you care enough about yourself and your career to learn enough to be proficient and eventually excellent. 4-year colleges, and in particular very good 4-year colleges and grad programs, work hard to get proficiency in what they think is relevant (with the above-mentioned proviso that they think a future in academia is more likely than statistics actually support) and open your eyes to the skills and factors for excellence.

I've known curious bright people who never got any 2 or 4 year degree or who got completely non-technical degrees who are world class programmers. They go to conferences, read journals, participate in technical professional development, etc.

If you assume just going to college is going to get you through, and not following up with conferences and journals and technical professional development, you're imposing a glass ceiling on yourself. You will not excel.

If you assume that your m@d l33t code hacking skills will get you through and that you don't need to care about algorithms and computer science topics writ large, you're imposing a glass ceiling on yourself. You will not excel.

If you assume that reading slashdot and a dozen more websites is an acceptable replacement for doing homework (reading actual tech journals, CS papers, etc), you're imposing a glass ceiling on yourself.

Grad students generally never survive to graduate degrees without understanding that, though not all succeed in the real world. A lot of 4-year students don't get that, even ones who went to good universities. Far too many 2-year university students and self taught people don't get it.

Put the video game down and go find out what researchers and cutting edge programmers are doing, what they see as the next hard problems, and find out what's going on which will be relevant to the work you're doing now, what you're going to be doing next year, and what you hope to be doing in your wildest dreams in two to five years. If you aren't actually going out and looking at the advanced stuff coming down the pipe your skills will erode over time, no matter how hot they are now. Widen your scope and look deeper.

Minor local damage??? It's a 400 kiloton equivalent airburst energy...

It will do 1 PSI overpressure (broken windows, etc) about 13 kilometers away from ground center point of explosion.

Right under the explosion, it will do about 2.5 PSI overpressure, and collapse relatively weak residential structures.

That energy level is going to kill people, if it's over inhabited areas. Not a lot of people - many or most directly under it would survive that overpressure level - but it will collapse things, and of a few things collapse people will die from the collapses.

Relevant calculations for blast overpressure:
http://www.nuclearweaponarchive.org/Nwfaq/Nfaq5.html Sect 5.6.2 Blast Damage and Injury

I bet I know what facility you're talking about.

I put the first Sun E10Ks in one of those, when they were pretty new. It took a while to communicate "No, we're serious, it's that many inches wide, that many inches deep, and weighs 1,600 pounds. Show me your engineering drawings on the raised floor, and I need to measure the elevator again."

Let's be clear - if you have hot-aisle cold-aisle with proper plenum separation, i.e. either walls or curtains blocking airflow from cold to hot other than via the racks and systems - then you can run all the air from the ceiling fine and it doesn't matter.

If you do not do that, you have to pump up the cold air pressure to jet it down so that it goes down into the cold aisles to floor level, and does not get sucked right back up into the return vents in the hot aisles in a loop up near the ceiling.

Once you pump it up with enough pressure to do that, you're at the same energy usage you would have had running it down into a raised floor plenum.

IF your datacenter has those hot aisle / cold aisle plenums with physical separation so it doesn't short circuit, concrete floor is fine. I've seen exactly two places that put them in, one of which screwed it up and was in fact in worse shape than before.

IF you just spray it down from above without separation, and you don't have enough cold air pressure, you are losing all the cooling efficiency circulating air without running it down to/through servers. Convection will make the hot aisle air rise, but not fast enough to win entirely.

Raised floor is not perfect. But it does have the advantage that if you cold aisle hot aisle with overhead exhaust, you can be reasonably sure the cold is done right just by managing the ventilated floor squares right and doing a little measuring. It is MUCH MUCH SIMPLER to make work with reasonable effectiveness and efficiency and much less likely to be grossly wrong.

If someone says "My datacenter floor is concrete for energy efficiency" and does not immediately then describe the curtains or plenum separations for hot/cold aisle, they are bullshitting you. They heard that it's better but don't know what will kill you in doing it that way, and they are probably going to get it wrong, orat least not get it right enough to save energy. If you find the right people and they're doing the proper plenum separation then go for it. If not, don't. The other people should use raised floors until the knowledge on plenum separation soaks out into the industry enough that they can do it right.

Lifetime - significantly better than Flash, 3 plus orders of magnitude more write/erase cycles before there's degradation.

Impact on overall computer heat & energy required to use - lower read power than Flash, no maintenance power (DRAM requires rewrite cycles as the bits decay)

Expected size - Initial model is 0.5 GB (512 MB) per chip. That's on a much larger fab process than current CPUs or DRAM though - expect that to increase rapidly once demand is established.

The teaser has a mixture of detail levels. They are probably still actively doing animation for parts of it, CGI production runs tend to end up like that.

There's one very sharp, in focus (mostly not motion blurred) frame of angry alien face at 1:39, after the armored marines come charging out of the lander, which is fully 1080 photorealistic, and could be a photo of a person if it wasn't for people's heads not being built quite that way.

That's what the movie's production team is *capable of*. What they do for most of the movie is unknown - some of the shots had no-better-than-good video game graphics. Some of them have far better. Hopefully the obviously deficient shots are rerendered at the level we see in the 1:39 face by the time they reach the screen. If not, they didn't spend enough money on the render farm (or take enough time), and that would be a pity.

Remote access is secure - SSH, RDP, decent VPNs are fine for remote administration.

If you don't trust the admin if you don't have them in your direct line of sight, why would you trust them if you're out of the room temporarily?

If you don't trust them when you're not looking over their shoulders, why do you trust them at all?

Either you trust them - and where they are sitting is irrelevant to that question - or you don't. If you don't trust them, fire them and get someone else you trust. If you don't trust them but think watching them in person makes it better, you're misjudging the situation and asking the wrong question.

Trust or no? If no, replace.

There are almost 200 customer delivered Tesla sedans on the road today (one owned by someone about 2 miles from my house, saw it being delivered a few weeks ago). They're all over the roads in Palo Alto if you happen to work here, though some of those are undoubtedly factory test drives.

It may take a while to get yours if you plunk down your $110,000 today, but it's a real shipping product.

That's not good enough for the likes like me.

I suppose all the UFS, NFS, VxFS, EXT2, EXT3, Reiser, et al filesystem bugs were not good enough for you either?

Anyone who thinks any filesystems are by nature entirely perfect is naive or too new to be pronouncing wisdom. They're pretty thoroughly debugged, but that's different than being perfect. The hardware/software interactions change over time, breaking design assumptions even in older FS code that's probably completely debugged. You get subtle hardware or firmware flaws in chips somewhere in the data transfer chain.

Things go wrong.

Either you understand this and prepare for it, or you're in trouble.

ZFS handles a lot of failures that other filesystems don't even realize happened, or can't recover from. It also has some new and somewhat bizarre failure modes of its own. You win some, you lose some.

Kono, that's not exactly true. I've also been on the zfs-discuss list for years, and there are at least two unfixed issues in it that I have seen.

First was the unplug-a-disk-keep-writing problem, where data kept being written into the cache but not committed to disk, so you got a consistent return and write verification at the application/OS level but on reboot the data's gone. This was first identified some years ago by Darren Dunham and posted to zfs-discuss by him and myself, using a E250 and the earliest public Solaris 10 betas. Nobody ever answered the question until very recently where someone else spotted the mechanism behind it (writes going to cache and being acknowledged, but not committed to disk). Still not fixed that I know of.

Second is a low level zpool / iscsi issue, where under some circumstances you can get into a situation where writes go into the cache and are acknowledged but reads get an older, "stuck in time" version of the data (we encountered this trying to manually overwrite an iscsi mounted Oracle RAC shared disk which was having Oracle problems, trying to dd zeros on top of the device... could dd all we wanted, either locally or remotely, but reading the disk kept stuck at the old Oracle ASM header). Nobody's exactly sure what's up with that one.

That said - UFS, VXFS, NFS, every filesystem in the world has had some bugs. Anyone who thinks otherwise is silly. Sysadmins and system architects have to plan for this...

Earlier work location:

BART round trip from my house to San Francisco, plus 30 min walk each trip (to station, from station to work): $8.60/day

Driving to SF: 22 miles, 1.0 gal gas each direction observed (currently $2.40/gal for midgrade), $4.00 toll, parking in SF free (when had handicapped tag) to $20/day. Car insurance and repair costs very close to $5/day. Total $13.80 - 33.80/day.

Cost of having car at home for social/shopping but commuting on BART: $13.80/day.

Current location:
Same as-crow-flies distance, but no direct transit links.

Auto to work and back, using other bridge, 22 miles (1.0 gal observed), $4.00 toll, parking free, $13.80 /day

Auto to work long way around w/o bridge 34 miles to / 22 miles back on bridge - $11.10 / day

Given that I can't walk long distances with heavy loads for shopping all the time, having a car is mandatory. Once I do that, the total cost to commute in the car to SF (urban center, pay parking) is more than riding transit, but to anywhere else is less than riding transit.

And there's a hefty fare hike coming from BART.

Oh, did I mention that about 80 cents per gallon comes out of my gas money and about 60 cents of that (plus a bunch of my sales tax money) ends up in BART and other local bus and transit systems?

When you're done factoring in subsidy from Cars to transit, the numbers no longer even vaguely favor transit. But that's apparently not PC.