There's also the argument that MJ's a gateway drug because of correlation - "you don't see many heroin users who didn't start with marijuana". My general reaction is "Look at all those heroin users who avoided marijuana because it's illegal and dangerous! (Oh there aren't any?)" (Actually there probably are some, people who got addicted to prescription opiates they started using for medical reasons, and switched to heroin because it's cheaper and because they can't get enough legally and weren't getting good medical support for getting out of the addiction.)

I did know one guy for whom marijuana actually was a gateway drug - first time he got high, in high school, he decided that it was good stuff and they'd been lying to him about all the reefer madness stuff, and figured he should see what else they'd been lying to him about.

Definitely. Depression and bipolar depression are widespread, and self-medication with alcohol and other drugs is fairly common. Some people are drunks or stoners or opiate abusers because they like it, but for a lot of people it's because they want to dull the pain or stress. For many of them, there are pharmaceuticals that could do a better job of managing depression or mania, but either they haven't gone to a psychiatrist because of stigmas about mental illness, or because their insurance doesn't cover it, or because they think they've got things sufficiently under control themselves, or because AA is keeping them sober, so they stick to the booze as their go-to self-medication. (Opiate abusers get other problems, because those are physically addictive; alcohol can be but it takes a lot more abuse to get there.)

There are other people who are tuned toward thrill-seeking, and like to hit the coke or whatever, but I've got less experience with them. And then there are other traditional reasons for drug use (mainly alcohol), such as boring jobs - farmers, video store clerks, etc. who can do their work just fine stoned.

It's actually a great comparison. Depression, both unipolar and bipolar, is quite common, and self-medication with alcohol and/or marijuana are quite common. One of my coworkers was bipolar and spent some time in the local hospital mental ward after a crisis, and said that most of the people there for alcohol detox were bipolar folks who'd gone off their meds (because meds are boring) and switched back to drinking until it caused them problems. Another coworker who's hypomanic said she used to need a couple of martinis or a joint to unwind in the evenings, but eventually went to a psychiatrist to get some better-tuned meds.

So yeah, some people are alcoholics or stoners because they like being drunk or stoned all the time, but for a lot of them it's really dealing with underlying mood problems that could also be addressed with prozac and its relatives. Cocaine's a bit different, but if you want to be manic-depressive and aren't that way naturally, it's a good substitute.

For Bitcoin, ASIC is the only way to go, but most of the interesting alternative coins are designed to be hard or impossible to build ASIC miners for. (They're also designed to be GPU-miner-hostile, but some of those have been worked around.) One of the tradeoffs with that is that CPU-only mining is botnet-friendly; it's harder to abuse botnet machines' GPUs (especially in cloud servers or routers that don't have GPUs.)

I avoid the whole problem by mining Dogecoin; it's close enough to no value that it's seldom worth stealing (though there was a botnet in the news recently that actually got $200K from mining it.)

No, generally a router has an inside and an outside, and sometimes a third port as a DMZ; you're thinking of a router with an ethernet hub attached, like many home routers. There are routers with more routed ports, and there are one-armed routers also, though that's less likely to be useful.

It's $35 plus shipping for the development board with the module soldered on it, so it's about the same as an Arduino; the $89 price was for two of them plus accessories like cables and power supplies. They're asking for not very much money to finish their software development, and the real question is whether their software is any good.

Sure, if you're just routing, and don't want to connect to various hardware I/O things, you can get a simpler board. But if you want to talk to sensors or build yourself a toaster controller or weather station or add lots of blinky lights or whatever, they're useful.

It's $20 for the module, $35 for a development board with the module soldered on it (which you'd almost certainly want), $45 for that AC power, cables, $50 for the board plus a spare module and pre-installed software, about $5-8 for shipping.

The reason the VPN connection went fast isn't that EEEVILLL Verizon was throttling the customer's Netflix connections by doing deep packet routing and didn't do that to the VPN. The pipes Netflix bought to deliver movies to their paying customers who use Verizon weren't big enough to carry all the demand, at least at the peering* point that customer's traffic went through, while the pipes they bought or peering they got for free were big enough to reach the VPN endpoint, and the VPN endpoint had bought enough bandwidth from their ISPs to get from there to their peering point with Verizon, so there was enough bandwidth on the whole route to carry the movie that way.

That's not to say that there aren't ISPs harassing particular content (there was at least one well-publicized case a few years ago of some telco ISP blocking VOIP, and of course most of the cable modem and some DSL providers block home web servers), but this ain't one of them.

(*Peering unfortunately means two different things here - it's giving each other service for free, and it's having BGP-managed interconnections, usually at the big internet exchange locations, to pass traffic, not necessarily for free.)

The problem isn't from Verizon's backbone to the customer - otherwise the VPN connection would have also been slow. It's the pipes from Netflix and their peering/transit providers to Verizon, which aren't big enough to handle all the Netflix customers on Verizon, at least in that region.

Also, what do you mean "has to set aside 8tb for sync"? Do you mean the ISP has to provide 8TB of some kind of storage hardware, or 8 terabits per second of traffic from somewhere to somewhere else? That's fairly huge, considering that most data connections aren't much bigger than 10 Gbps per wavelength (some carriers use 40, but it's not usually price-competitive.)

Netflix either buys bandwidth or convinces ISPs to give it to them for free as peering. Its peers / transit providers are either big enough to peer with Verizon for free, or buy bandwidth from Verizon, and they don't have fat enough connections to Verizon to carry all of Netflix's traffic to this customer, at least at the peering point that Netflix's traffic uses to reach that customer (Netflix may have enough bandwidth to reach VZ customers in the other part of the country, e.g. LA peering is full while Seattle or DC isn't.)

The customer's VPN provider may very well provide free/cheap connectivity to Netflix (or at least fast enough), maybe even one of the providers Netflix buys from, because they're cheap. And they may also get it from somebody who has good bandwidth to Verizon. But peering's just peering - it's not transit.

(First of all, to reply to the parent article, the test isn't for people graduating from college, it's for people in high school who will get to use the results to place out of courses in college. In my case that meant I could start more advanced calculus classes a year early, which was really useful, and got some extra credits for biology that didn't affect anything but probably looked good, and if I'd been at a college where tuition prices were by the course instead of the semester, it would have probably saved me some money.)

So it's nice that 39,000 students had enough high-level CS courses in high school that they were able to take the test, but that's a pretty small fraction of the number of kids entering college, even just in the STEM fields, and it's also limited to those high schools that had a good enough CS program to make it worth taking the test, so the statistics are probably not all that representative*.

(*And the fact that I misspelled "representative" in the title doesn't mean I'm bad at English; I ran out of characters in the title box. :-)

Because nobody out there has "customers who bought X also bought Y and Z" features! (But maybe this one will let Apple take some more market share from Amazon.)

The Russians didn't shoot down that plane. Ukrainian separatists did, using missiles they got from the Russians.

And it's not like the US hasn't accidentally shot down civilian aircraft before, if you remember that Iranian plane the USS Vincennes shot down.