Rail currently handles 40% of all goods shipping in the U.S. Before NAFTA this figure was much higher, close to 75%, but it has dropped because of the difficulty rail has crossing borders. So you're wrong; rail is a very efficient method for massive movement of goods that has lost ground to more expensive truck freight because of political restrictions on its use.

Paris-Marseille, TGV, 489 miles (783km). 163mph avg, 3 hr.

I just checked, and they currently have P-M specials for 83 euro. So that's roughly the same as your $120 special.

While the French government owns the rail network and makes major construction, TGV is operationally self sustaining.

There's no reason rail wouldn't work here. As someone who regularly commutes about the same distance SF-LA I'd love to have the legroom of first class air for coach fare, a nice quiet cabin, electrical outlets and internet access, a smooth ride, and a cafeteria car. Above all, I couldn't be happier if I never had to stand in an endless security line again or sit around for hours because there's a blizzard in Colorado and the entire system is backed up.

They won't be able to use the full feature set though. A framework or library, like OpenCL, which wants to map GPU memory into the process address space will likely not be full featured on IA32. It likely won't find a large enough hole in the virtual address space to fit a 1-2GB region, or even a 512M. So the compatibility mode version of these frameworks will either exchange data using a buffered DMA model, a remapped window, or only use a small portion of video memory - say 128M. The compatibility mode versions will by necessity be limited functionality, and since much of Win7 (like OS X) leverages the GPU for processing these limitations will percolate to other parts of the system.

Easy and usable are orthogonal really, and people often mix and match terms like these to express what they like. They might say it's "easy" when what they really mean is that it's "uniform". Subjective experience is often difficult to explain because it's inherently non-verbal.

One thing I really like about Apple laptops is 1) the speed at which they're ready to use (about 250ms after opening the lid if the battery wasn't allowed to drain totally), and 2) how wifi is managed. It'll look for SSID announcements, collect a list, sort the list by a) whether I've used the network before, and b) signal strength. It then picks the top candidate and tries to connect. It usually takes a second or two until it's configured from opening the lid. If I want a different network I pull down the list and pick one. What else is needed? The failure mode is it may pick the wrong network if there are multiple choices and signals are variable, or it may miss an SSID broadcast. So occasionally when I'm in the front of the house it may pick the access point in the back, or the opposite. It's not a big deal and frankly isn't all that common. My wife recently switched from Windows, and I don't know how often I had to configure the wifi vendor utility on her Dell laptop to select the right SSID for her location. Not to mention reenter passwords every month or two when it decided to go amnesic.

It doesn't matter what hardware is used, it still works the same. In fact I don't even know what card is in my laptop, and couldn't care less. My coworkers' laptops work exactly the same, and I don't know what hardware they have either - and it doesn't matter. It works the same.

Similarly, the virtual desktop (Spaces) works the same regardless of whether I have an nvidia card in my laptop and an ATI card in my Mac Pro. Or my wife's 13" MacBook. There is no vendor specific component here, it's not a driver feature, or a value add. It's simply basic common functionality. To be honest I don't even know what video card my 2006 MBP has (ATI I guess) - and it doesn't matter. I couldn't care less; it's only important in the Windows world where I have to locate the right driver with the feature set I want, whenever a Windows patch would cause trouble because the video driver was too old. In my Mac Pro I have two ATI boards (driving three monitors), but I could have one ATI and one nvidia board, and it would still work the same. How does Windows behave with a mix of video card brands?

Finally, from a purely engineering standpoint, I hate the unified monolithic Windows registry. On OS X I can go pull a plist file for some application from a backup, but with the Windows registry it's all or nothing. Of course backups are now reasonably uniform with Time Machine as well. (All the machines in my house backup using TM to a NAS. A full restore requires nothing more than a blank system install, no third-party software needed.)

Some people really seem to genuinely want their system to work like a Japanese-market cell phone. The more controls (no matter how pointless) the better, two utilities to do the same is better than one, each system should work differently, etc. I don't know if that makes them feel like some arcane loremaster, or like they're doing something useful, or what. But they'd be much better off actually learning something that creates real value, and when they do they're not going to want to spend half a day getting a video card driver to work reliably - just so they can finally get to do whatever it really is they want to get done.

This isn't feasible if the TPM is used to block encrypt the drive with a host key. It will only work with one specific CPU.

No, he doesn't want to mention it because it detracts from his message. The fact that it runs Linux is irrelevant; that's not a selling point but merely an implementation detail. I work with embedded Linux devices - TVs and Blu-Ray players currently - and the fact that they run Linux is about as relevant as who made the chips or which factory assembled the boards. The people who buy these devices care about what they do, not how they do it; discussing the how part is totally irrelevant. If you try to sell a "Linux-based TV" people will wonder why they should want a Linux-based one instead of a regular TV - and in fact might be led to believe they're buying an all-in-one Linux computer and TV. Which they don't really want. They just want a TV so that's what you will sell them, and you will point out its superior TV functionality. Unless they ask you don't mention implementation details.

This is not entirely true. The loans are due before the money is needed for pay-out. While borrowing from SS is a bad precedent, it's not quite so bad that it's scheduled to go insolvent. It would only be insolvent with the funds on hand today, excluding loan repayments. Claiming that SS is going insolvent also implies that somehow the treasury doesn't intend to pay it back. It has paid it back so far, so there is no reason to assume otherwise in the future. Finally, SS is getting the same interest as anyone else the treasury borrows from, so is getting some return, with just about zero risk. SS, for instance, wasn't gutted by the recent financial troubles and that's making this particular 'investment' look pretty good right now. Certainly beats the pants off my 401(k)'s.

I was denied coverage some 6 years ago because of an "undiagnosed rash". I still have no idea why they decided I had that, or why a rash would be a reason to deny insurance, but I suspect they probably had a copy of old records since me and my wife had been covered by the same insurance company previously. She was denied because she had quit smoking too recently (3 or 5 years) - she claimed non-smoker, but apparently her old records said she smoked. There was no request for clarification, or an interview, or adjusting the rates. They just refused to insure us. So now if I we were to apply for an individual policy we'd have to disclose we've been denied insurance previously, which means we're probably uninsurable.

In the U.S., my private insurance which is fairly mid-tier only covers in-network facilities. If I try to skip a queue by going out-of-network I pay 50% out of pocket. The insurance doesn't cover elective procedures or experimental ones like chiropractors and acupuncture. I recently got running shoe inserts by a sports orthopedist, which I paid for 100% out of pocket. They'd never pay for any sports medicine, like LT testing or bike fitting, or anything like that. When I had shoulder bursitis and needed surgery to remove a bone spur in the cuff area I had to wait 7 weeks.

And then on top of that I usually get bills for things that should be covered. I once got a $700 bill for a standard blood test. It can take a few months for the insurance company and provider to figure out how much to bill and who pays. It's utterly byzantine, and the bureaucracy is impenetrable for an outsider like me. Basically, the system here is utter garbage.

And I'm a principal engineer who makes $145k/yr not counting stock options. I don't have crap insurance - it has high limits and in the end does pay. Shitty low-end plans have low limits and frequently drop you entirely if you get sick and lose your job (meaning the group administrator isn't likely to retaliate since you're no longer an employee).

It's actually pretty hard to get kids to run even 35-45min, because they lack the ability to pace themselves. They slow down only as they burn out, and after 20min they're unable to continue. An HRM with a zone alarm is just the right tool if you want them to pace themselves.

However, unless they're training for athletic performance and to compete I'm not sure why it would matter.

This is exactly what it's for; the chip identifies the passport number, which is then looked up in a database. If the chip isn't responding to the scanner the immigration officer will simply type in the passport number manually from the print.

RFID tags and scanners come with various degress of security, for instance a 'secure' RFID doesn't even respond to a scanner it doesn't recognize. It simply stays silent. The tags on ID cards are generally not secure - this is a cost issue mainly as secure tags and scanners, and the infrastructure for key management adds significant expense. The more secure, the more proprietary and single-vendor.

The problem is that by pairing a scanner a webcam it's possible to create a database of federal employees and their identifying tags. This in turns allows anyone with a copy of the database to easily track whether people entering or exiting a room for instance are federal law enforcement, or anybody else in the database for that matter. As you can imagine federal law enforcement would rather not this be public information, but do have a need to be able to identify themselves as law enforcement so have to carry ID. They might as well pin it to their chest if it has an unsecure RFID tag.

Obviously it also allows more elaborate tracking and analysis, by for instance seeing which law enforcement officers often come together or stay at the same hotel.

It would take significant changes to the language to allow removal of pointers and arrays. Most current attempts to do so simply punt all the details onto the underlying allocator, which doesn't have sufficient information to do a good job. As an example, an ethernet I/O buffer might need to be allocated in a memory region mapped onto the device that implement the interface chosen by IP as the output interface - cached by TCP for a connection. And some ethernet devices might have such memory mapped buffer areas; others don't. Implementing this in the allocator requires carrying a lot of random baggage (context) up and down the call tree. It easily results in difficult to resolve catch-22 situations that will then impose design restrictions that actually prevent simple and straightforward implementations. And then you still have layout restrictions like how you want the IP header aligned on top of the PHY framing. In the end you'll likely find pointers and arrays used to implement containers that provide all the necessary functionality and that can be mocked and unit tested really is the better solution.

Modules and GC was no big loss since they're not all that useful for system programming; C++ would typically be used to _implement_ things like allocators and garbage collectors. Modules is neat but could often actually restrict code reuse. But concepts would have been extremely useful for code reuse, where templates result in excessively complicated constructs. Templates are commonly avoided for anything other than basic patterns and containers for this reason, which is probably advisable in many circumstances. If you do get it working in one circumstance it's not self-documenting when it will or won't work, i.e. what's expected from the template parameters; the functions/interfaces they need to implement. So a template used to wrap say a PCI ethernet driver might behave erratically or even fail to compile when used to wrap a USB device. Even programmers who understand templates and can effectively debug code factored with them often don't have a solid understanding of the impact on runtime footprint - when the resulting code emitted bloats beyond all sanity, and when it's only a symbol and namespace game. The code itself doesn't express this.

Pair composing is extremely common.

I remember learning programming in the late 70s (7 or 8th grade probably) by sitting with a friend next to a computer and working together on programs. Both BASIC and 6502 assembler on the Apple II. This wasn't part of any training course or intentional teaching scheme; it was simply a necessity since there were two neophyte programmers and one computer. I think it's a good experience and an effective teaching aid. I also think it's completely useless and counterproductive for more than getting started. Once they can write working programs (not necessarily pretty, but functional) collaboration is more useful. Get used to reading svn commit mail and see what changes others make, review them, and look for improvements. Also teach them to look over and review their own changes like if someone else had made them.

I suspect it's more that intelligent people are able to abstractly consider themselves and their own behavior, and accept they have a neurological disability. People who are purely reactive to their environment and don't proactively "push their own cart" so to speak are less likely to reason around their own behavior or ask themselves why they do what they do. The cause of schizophrenia (i.e. the inability to distinguish fantasy and their own speculative thoughts from reality) likely has nothing to do with this; it's just that sufficiently intelligent people don't let it become a problem when they're aware of their own tendencies. Also, once you start hallucinating - hearing voices, seeing things - your case is probably so severe it can't be self managed and requires anti-psychotics. If you don't take anti-psychotic medication and continue experiencing hallucinations your brain will soon adapt and wire itself to respond to them as part of your environment. This makes gaining insight into your own condition progressively more difficult, and as time proceeds the condition gets more and more difficult to treat, since medications can't unwire long-term adaptation.