Except this time, Netscape is metaphorically Microsoft's product, and IE smells like fresh, hot gingerbread. Mmmmm!
Take away Google, and the streets will riot, Take away Bing, and... er... um... someone might eventually notice. Maybe.
Hotmail? Is that actually used for anything besides MSDN SSO credentials anymore?
IE? (rolling on the floor, gasping for breath amidst near-lethal amounts of laughter). People with *Windows Mobile* didn't even use Internet Explorer before Android came out. We used Opera, and paid for it, because Pocket IE sucked like a whore with braces. I'm sure Microsoft did a much better job with IE on WP7... but then again, a chimp with a Commodore 64 and a pirated copy of GEOS could probably improve upon PIE in its WM6.x incarnation.
MS Office Suite? Meh. Apps to view and edit word/excel docs are free and abundant. Outlook still can't do adhoc aliases properly, and Android can be induced to lie about its authentication capabilities so you won't have to indulge your company's IT department's wet dreams by entering a 16-digit passcode before it allows you to answer an incoming phone call.
What Microsoft REALLY needs to do a better job of is tracking mouse movement (specifically, acceleration, ballistics, etc) to do a better job of discerning intent when you go to grab something... and a better job of adaptively figuring out over time whether its assumptions about your intent are right or wrong. It really seems like every new version of WIndows leaves me fighting and frustrated with it a tiny bit more.
Here's a real-world concrete example. Suppose the mouse pointer is approaching the right edge of a window that's maximized to the left panel of a multi-monitor setup. The mouse pointer slows down, and seems to also be approaching the scrollbar. The left button gets pressed, and the mouse moves in a direction that's mostly upwards. Well, except the pointer overshot the edge a bit, and the left click technically occurred 2 pixels into the window on the monitor adjacent to the right. Taken in isolation, Windows has no real choice but to assume the user meant to click the pixel on the other monitor even though it contextually makes no sense. But combined with the observed ballistics (slowing down, slight arc towards the scrollbar, motion after left-click that makes more sense as a scroll-gesture than a... well... meaningless gesture), it's obvious what the user meant to click. And for the most part, Windows, seems to be completely oblivious to it.
Now, for a counter-example: trying to select text without adjacent whitespace. For me, Windows (Word and Outlook in particular) NEVER seems to get this right. I'll click at the right starting point, letting go and starting over if I'm not happy with it. Then I'll start highlighting. But way too often, it'll stop selecting a character or two short of where I want. If I keep moving the mouse, it'll grudgingly select the remaining characters... but feels compelled to ALSO ignore my hard initial-selection work and expand the other end of the selection too. Dammit. I'll then spend the next 10 seconds fighting with it trying to select the text I REALLY want. Half the time, I'm forced to give up, let it select the damn adjacent whitespace, and edit it away after I paste. It annoys me to no end.
In the end, it feels like Windows has simultaneously gotten worse in two directions. It forces its opinion on me without learning from its mistakes or giving me the option to beat it into submission so it quits interfering, and simultaneously forces selection with almost single-pixel precision to make increasingly-dense window gadgets work. I'll admit that Java is even worse in this regard, and Linux (or at least Gnome/Compiz) doesn't seem to be any better, but it's still annoying as hell.
You know, if I could go back in time and try to change somebody's mind, I'd make it my mission to try and convince the French to define the meter such that 24mm were precisely equal to one imperial inch. Why? Because then, instead of having two sets of tools requiring two entirely different sets of screws and bolts, we'd only need one, regardless of whether you prefer metric or imperial, because measurements in either system would have precise, nice, exact equivalents in the other. 1 inch? 24mm. 3/4 inch? 18mm. 1/2 inch? 12mm. 1/4 inch? 6mm. 5/8 inch? 15mm. 7/16 inch? OK, not as pretty, but still a reasonable 10.5mm. 23/32 inch? Ugly, but still-tolerable 17.25mm (compared to an abomination like 18.25625mm). You'd basically have to go to 64ths of an inch before you ever had to deal with fractional millimeters besides.25mm,.5mm, or.75mm.
OK, so it wouldn't have been 1/10000000th the distance from the equator to the poles... but it turns out, they were wrong about the distance, so the meter's length ended up being as arbitrary as the length of some ancient king's foot anyway.
Check eBay, it's littered with sub-600MHz Android tablets with barely enough flash to hold 1.6, barely enough ram to boot, no Android Market, and likely to spend most of their lives as... digital picture frames, because most users aren't masochistic enough to suffer with them for more than a few days once the novelty wears off.
Don't write them off, though. They might be useless as general-purpose Android devices, but they're cheap enough to use as single-purpose devices (home theater remotes, family photo albums, home automation controllers, interactive cookbooks, etc) whose ultimate use just happens to have not been carved in stone (etched in silicon?) at the factory.
Nevertheless, if you really want a general-purpose Android tablet to play with NOW, don't touch ANYTHING that doesn't ship with 2.1 or better, have a 1GHz or faster CPU, enough battery life to actively use for 3-4 hours, and still have enough of a charge after laying ("off") on a table or sofa cushion for a day or two to use for 5-10 minutes before you HAVE to put it back on the charger. Try to find one that has 1280x800 or better resolution, because that's the magic point where you can almost read two O'Reilly-sized pages side by side from a pdf file. Not coincidentally, demanding better than 480x800 ALSO happens to weed out most of the tablets that are too slow/limited to be worth bothering with right now.
Oh, and just to warn everyone... don't buy a Galaxy Tab unless you're absolutely delighted with it as it exists RIGHT NOW, because Samsung has already fucked millions of customers who made the mistake of buying a Captivate, Vibrant, Epic4G, Fascinate, or Mesmerize. Maybe it's Samsung's fault, maybe it's the carriers' fault, but either way, if you buy a Galaxy Tab, it's coming from Sprint, Verizon, AT&T, Verizon, or some regional carrier. If they don't even care enough about their millions of PHONE customers to upgrade them to Froyo and fix the goddamn dysfunctional GPS, does anyone really think they're going to pay more attention to a few hundred thousand tablet purchasers?
> Mobile Windows didn't have such glaring problems with malware stealing from the user.
That's mostly because statistically, there weren't enough Windows Mobile users (or PalmOS users, or Symbian users, for that matter) to be worth the time of organized crime.
The problem with Android isn't the fact that the source is available to peruse, it's the fact that manufacturers and American carriers do their best to make upgrades as difficult as possible despite Android's open-source Linux roots. An exploit like this barely gets a yawn from Nexus One users, because someone will update it before any real exploits based on this ever become a problem. In contrast, owners of American Samsung Galaxy S phones will be shitting bricks, because we're still waiting for a fucking kernel that works with Froyo. Or at least leaked CDMA loadable kernel modules compatible with a 2.6.32 kernel so we can build our own without losing basically all the hardware drivers it needs to work properly.
Sidetrip: Unlike Windows, Linux makes no effort to maintain a stable ABI between versions. Simplified a bit, this basically means that a loadable kernel module (the Linux analog to a hardware driver) that's built for a 2.6.29 kernel will probably crash and burn on a 2.6.32 kernel. The official Linux party line is that it makes it harder for manufacturers to keep drivers proprietary, and motivates vendors to release source for their drivers so it can be automatically rebuilt for each new kernel release. The cold American consumer reality is that the Android Emperor is nude. The Nexus S can't do 4G on T-Mobile, is fundamentally incompatible with Sprint and Verizon, and AT&T's slow, capped, expensive 3G isn't even a real option. We're stuck with an allegedly-open operating system inextricably bound to hardware that's more locked down and proprietary than an iPhone, and all we can really do is hope some of Linux's core developers also own Android phones and are starting to really, really feel some of the ABI pain themselves on a daily basis.
Put another way, here's a more technical summary of the problem:
* Samsung has released source to its kernel and loadable kernel module drivers, but the LKM source won't build against any known 2.6.32 kernel due to missing dependencies.
* The.ko modules themselves were built against the ABI of a specific build of 2.6.29 that changed enough with 2.6.32 for most of them to crash and burn if you try using them with a 2.6.32 kernel.
* Froyo and Gingerbread have dependencies on the 2.6.32 kernel. You can cobble together a FrankenBuild that sort of works with a 2.6.29 kernel, but it'll never be a True Froyo/Gingerbread, and will always have bugs hidden below the surface veneer.
Metaphorically, an American Samsung Galaxy S trying to run Froyo is kind of like a laptop that shipped with Windows 98 and a winmodem. The unfortunate user upgrades it to XP himself, then discovers that the winmodem only has drivers for Win98. Through some miracle, the winmodem drivers have their "source" released, but that source requires a thirdparty library called LunexantProprietaryLib that isn't included, and won't build without it. After lots of hacking, the user manages to cobble together drivers that will allow the modem to limp along at 9600 baud by pretending it's an older version of the chipset, but getting it to do 56k without official drivers is hopeless. And if, by some miracle of god, a never-released copy of drivers for XP get leaked despite the determination of the manufacturer to keep it unavailable through the perverse logic that fucking their customers will somehow encourage them to buy a newer model from the same company that screwed them less than a year earlier (instead of buying one made by just about ANYBODY else), the user discovers that the drivers needed for 3D acceleration have the same problem as the Winmodem, and it's back to square one.
What Google really needs to do is define an ABI thunking layer and require that any and all device drivers
Not to mention, passengers in a train have self-help options in the event of catastrophe that passengers in a plane lack. They might or might not be useful, but as a practical matter, if you're on a 747 that gets blown in half 3 miles up, and by virtue of being in the rear tail section that gets blown away, well... you'll get to have 3-5 terrifying additional minutes of life that your fellow passengers didn't. Ultimately, though, you're as screwed as the rest of them. In contrast, if you're a passenger on a train that gets blown in half and your part skids away from the rest... well, you might actually live to tell about it.
There's also the fact that American passenger trains are slightly unique. By law, they have to be rolling bank vaults capable of surviving a head-on collision with a mile-long freight train. 99.9% of the time, that's stupid and bad, but when things like explosives get involved, it *does* tend to come in handy.
The main problem with recent (past year or two) HSR proposals is that they take a generally good thing, then ruin it by demanding an expensive level of immediate perfection that will ultimately keep it from ever really succeeding by virtue of both cost and limited scope. Take Florida. By all means, build the tracks between Orlando and Tampa to 220mph HSR standards. They're going to be around for a really long time, and it's worthwhile to do that particular job right the first time around. It won't raise the cost by much, and it'll seriously reduce the construction misery 25 years from now when the higher-grade of trackwork is *really* needed. On the other hand, fuck the electrification and trainsets that do 180mph, but can ONLY run on their dedicated, purpose-built tracks like a big amusement park ride. By all means, design the tracks to be electrified *later*, but for now, stick with diesel. You *don't* electrify a rail line that has one train per hour in each direction. It's economically insane. Not even the former Soviet Union was *that* crazy (at least, on a large scale; I'm sure there were a few tracks here and there that were electrified for the sake of politics, but I'm pretty sure the Trans-Siberian Railroad STILL isn't fully-electrified all the way to the Pacific). Instead, connect the new track (which HAS to be built for any kind of meaningful passenger service to Tampa from either Orlando or Miami, because the last 20 miles of existing track into Tampa ARE very, very intensively used by CSX for freight) to the existing CSX tracks from Auburndale down to West Palm Beach (which are almost commercially useless for freight), build a second, new track alongside it, and use the cash that WOULD have gone to run semi-useless 180mph trans back and forth between Tampa and Orlando to build a useful passenger rail network with 100-110mph average speeds between Miami, Orlando, and Tampa... maybe even Jacksonville a year or five later. FDOT has studies dating back to before most of us were born that have credibly concluded that 80-110mph passenger service in Florida wouldn't just pay for itself... it would make a real, honest to god outright *profit*. Those same studies invariably conclude that "real" HSR would hemorrhage money forever, mostly because the interest on the construction bonds used to finance the staggering up-front construction costs would crush any hope of it ever achieving financial viability. Florida's politicians need to realize... trainsets are a 20-30 year investment, but trackbed and right of way are forever. It's easy to electrify an existing rail line 25 years later. It's damn hard and expensive to try and straighten out curves after the fact.
Put another way:
* No-compromise "True" HSR between only Orlando and Tampa: crushingly expensive and minimally-useful.
* No-compromise "true" HSR between Miami, Orlando, and Tampa, built NOW: very, very nice... but it would bankrupt FDOT, and would more or less forever write off service to Jacksonville and southwest Florida, let alone Tallahassee (Tallahassee's problematic, just because it's so freakin'
> It has a presence, yes... but doing "just fine"? The iPhone and Androids each have more units in the channel than WP7 has in-channel and > activated *combined*. This is in spite of the fact that WinMo (in various incarnations) have been for sale for (almost) a decade.
They don't exist on Sprint and Verizon, yet. In the United States, that makes them commercially irrelevant because T-Mobile is tiny, and NOBODY voluntarily uses AT&T unless they're shackled to a pre-Verizon iPhone.
Likewise, Microsoft's success or failure with Windows Mobile is largely irrelevant. I'm sure there are 3 or 4 Windows Mobile users left who haven't jumped to Android or IOS, but I don't know them personally. The fact that Windows Phone is metaphorically "Sidekick5dotnet" and doesn't run Windows Mobile software makes almost any link between them irrelevant as well.
Still, as others have noticed, Microsoft can afford to throw monkey wrenches into the plans of others for a really, really long time. In the long run, Microsoft probably WILL recover a chunk of its marketshare with specific respect to the enterprise market if they ever get their act together and make Windows Phone integrate *seamlessly* with Exchange Server and Microsoft's management infrastructure.
There's one specific case where it's easy to replicate a visible effect: GSM noise. It drives air traffic controllers insane, because they get hammered by it all day. The irony? It's only really a problem when the GSM radio lies within a few feet of the microphone. Yup, that's right folks... basically all of the GSM noise that assaults ATC daily comes from the (powered up) Blackberries of PILOTS in the cabin. Even then, the solution is easy: use Sprint or Verizon phones, because CDMA doesn't cause the problem (technically, a GSM phone operating exclusively in UMTS mode wouldn't, either... but I don't think it's actually possible to selectively disable only legacy TDMA-based GSM on a Blackberry or iPhone and force it to use ONLY WCDMA-based UMTS).
The ban is almost entirely a matter of regulatory inertia and risk of lawsuits. Since the bans are universal, no airline wants to risk paying higher damages in a lawsuit if there's a crash and someone is able to convince a jury that their policy of allowing electronic devices theoretically increased their liability by even.000001%. On the other hand, if airlines could install picocells that made it impossible to connect to a carrier's towers, but enabled them to collect $1/minute roaming charges, you'd see any hint of a ban eliminated within a matter of days.
They're probably generating a 22KHz sine wave and running it through a transformer to boost the voltage from ~0.7 volts (more or less what you'd get if you drove the output DAC at the highest power output it could sustain with a non-fully-charged battery without clipping) to 3.3 or 5.0 volts. Yeah, you can do the same thing with about $10 worth of active components, but if you can easily generate a real sine wave to start with, you can basically replace the solid-state charge pump with a transformer, two diodes, and a capacitor.
As far as parasitic power goes (a fancy way of saying, "Drawing power from the data bus"), one way I know of is to drive your data line at two voltages... say, 6v or less for 0, 9v or higher for 1. At the other end, use a linear voltage regulator to convert it back down to 5.0v for power, and connect it through a resistor to an optoisolator whose other end is sampled by a GPIO pin. In real life, it's a little more complicated, because the capacitors you need would affect the signal, but I'm pretty sure the general idea would work (at least, for slow data rates).
My favorite back-channel approach to overloading power lines for data (in scenarios where you aren't trying to cover distances where voltage drops are an issue) is to add a diode bridge, and sample one output leg of it with an i/o pin. For example, suppose you have a microcontroller that has two pins: one for 5v DC, and one for ground. Now, suppose you need a way to convey a single bit of information to that chip in situations that really only matter during development or debugging. By sensing the output of one of the diode bridge's leg, you can tell whether the user connected the power normally or backwards, and act accordingly.
My own experiments have been Android-related (exploring the headphone jack as a gamepad input), but I came up with two obvious configurations for using a jack with ground, mic (input), left audio, and right audio:
* SPI (.7v logic): left audio = clock; right audio = master out, slave in (phone is master, obviously); mic = master in, slave out. Ground=ground. ~8-14 months from now, my Epic4G is going to be my Bioloid's new backpack-mounted controller:-)
* UART. left audio =.7v sine wave for power, right audio = TxD, mic = RxD, ground = ground. Obviously we're talking about nonstandard voltage levels and not RS232.
Other possibilities:
power circuit using left audio sinewave, use right audio to modulate infrared LED and make a cheap remote control.
bitbang NTSC s-video chroma & luma signals via right and left audio connectors... or mono audio + composite video. If a 20MHz 8-bit CPU can pull off color video better than most late-1970s videogames, an Android phone with 1GHz CPU hijacked into single-user Linux mode devoting its full attention to bitbanging the i/o should be able to pull off graphics at least as well as the Atari 2600. OK, pretty useless, but damn it would make a cool video on Youtube;-)
If the left audio, right audio, and mic pins are really GPIO with data direction registers somewhere, you could hack a decent 3-channel logic analyzer (or 2 channels + trigger). If the phone has a crossbar switch between the USB port and the rest of the phone, you might even be able to hijack the two pins normally used for USB data and get two more inputs. Add a bit more hardware, like a FTDI usb bridge chip with FIFO and bitbang-mode capabilities, and an old Android phone (or iPhone) capable of activating USB-OTG would make a totally kick-ass pocket logic analyzer.
Argh. Accidentally deleted an entire section and didn't notice until it was too late.
As far as how incoming calls are handled when there's an active data session, it depends entirely on the phone. The CDMA2000 standard itself has a perfectly good way for phones to poll for incoming calls, text messages, and voicemail notifications when there's an active EVDO data session. The problem is that historically, most phones get it wrong and totally fuck up the implementation.
Example: Sprint PPC-6700 and its evil twin, the Verizon xv6700. The original radio firmware took a long time to switch between EVDO and CDMA voice, so Microsoft (or whomever was customizing Windows Mobile 5 for the phone) decided to handle incoming voice calls during EVDO data sessions by pre-emptively breaking the data connection, then lamely asking (too late) whether you wanted to end the data session to take the call. Angry users eventually figured out how to hack the phone to make it refrain from breaking the connection until after you told it whether you wanted to take the call. Unfortunately, that hack caused lots of calls to end up going to voicemail at the last second, because the phone couldn't switch to CDMA voice mode in time to take the call. You'd get the alert, acknowledge that you wanted to take the call, and have it go to voicemail anway. The arms race continued, and we figured out how to make the phone poll the tower for incoming calls more frequently. It only increased the time we had to accept the call by about a half second... but that half second was enough to keep the majority of calls from falling through the crack and going to voicemail.
Enter the Sprint Hero. In its virgin 1.5 glory, it didn't even bother to ask whether you wanted to take the incoming call... if you had an active data connection, the call would silently go straight to voicemail. Worse, you wouldn't even find out about it until you laid off the data use long enough for the phone to close the EVDO session and go back into 1xRTT mode. After lots of hatred from angry customers, Sprint made it less bad sometime around Thanksgiving so calls still went straight to voicemail, but at least the phone properly polled for new text messages and voicemail notifications during active EVDO data sessions. As far as I know, Sprint's official ROMs *never* fixed the problem entirely, but I quit caring the day I discovered AOSP anyway (which did, in fact, fix the problem).
The Epic4G mostly seems to handle the situation properly... if you get an incoming voice call while you have an active EVDO session, it notifies you immediately, but it doesn't break the data connection unless you actually take the call. And if your data session is via wifi or wimax, it continues to work.
Truth be told, I'll be shocked if the iPhone DOESN'T handle this situation gracefully. Where things get sticky is how it will handle the opposite scenario (user tries to make outgoing phone call while running one or more networked apps). I suspect American apps will handle it better than European apps, just because EDGE and GPRS are facts of life in the US, but you can allegedly go kayaking through the fjords in northern Scandinavia, take pictures of a real reindeer, and share them with your friends 8,000 miles away via UMTS, even if the cell site HAS to be solar-powered because the nearest power line is 100km away;-)
Here's the deal: officially, CDMA2000-EVDO can't simultaneously do voice and data. Specifically, it's impossible to simultaneously have an EV-DO data session and circuit-switched CDMA voice call using the equipment Verizon and Sprint purchased from Qualcomm. However, the line has gotten a little blurred lately, especially with rooted Android phones (and almost certainly, jailbroken CDMA iPhones by summer). Five years ago, the answer to "Can you do simultaneous voice and data with EVDO" was unambiguously "No, it's impossible." Now, it's more like, "well... it depends. Do you have a rooted/jailbroken phone? Are you OK with spending another $10-25/month for thirdparty VoIP service? Good. In that case, the answer is, "Yes, kind of."
There are a few projects to enable users with thirdparty VoIP service and rooted Android phones to use it with their phones. Basically, here's how it works:
* You sign up for VoIP voice service, get a phone number, and configure it to forward incoming calls to your cell phone's number.
* You run an app on your phone that intercepts "incoming call" notifications when there's an active EVDO data session.
* When the app sees an incoming call, it checks to see whether it's from the VoIP number. If it is, it quickly establishes its own VoIP session over the EVDO data connection, launches the VoIP phone app, and lets you take the call without losing your data session.
The catch is, the apps (AFAIK) can't do traffic-shaping on your local network connection, so it's up to you to make sure that nothing else running on the phone saturates it and causes it to drop too many UDP packets. The last I checked, they were still kind of raw, unpolished, and required Android phones that were rooted to really work reliably (or at all)... but I don't see any obvious insurmountable reason why they couldn't work just as well with a jailbroken Verizon iPhone.
> If someone can sign your name on a paper and send it by mail you'd be fucked to....
Actually, no. You could legitimately argue (in court, if necessary) that your signature was forged. Forgery is so common, assertions of it in court are almost automatically accepted by juries as credible unless the party claiming it's legitimate can bend over backwards and demonstrate (through supporting evidence, like driver's license data, video surveillance footage showing the individual perform the transaction, etc) overwhelming evidence that it's legitimate.
Smart card-based certificates upset that delicate balance of power. They don't prove that it was signed by you, but they do prove (almost beyond doubt) that something was signed by someone with physical possession of your card/cert and knowledge of its security code. Thus, they instantly shift the issue from claims by the victim that his signature was forged (something that's happened throughout human history, is commonplace, and an easy defense for consumers to successfully raise in court) to claims by the banks that you were negligent in your handling of the certificate and/or its security code. As a consumer, you have basically no duty to prevent someone else from forging your signature, because you can't. And the scenarios where banks could claim you were negligent would be almost impossible for them to prove. In contrast, with the cert/card, if anything goes wrong, banks have a MUCH easier time of shifting liability to you, the consumer.
You could argue that a similar situation exists with ATM cards, but ATMs have an advantage (for consumers) that internet transactions don't -- pervasive video surveillance. If a criminal coerces you to give up your PIN code, it's likely to be pretty easy to prove his involvement and demonstrate coercion. If the criminal is out of view, but the victim claims otherwise, the bank's in an awkward position. If the bank were to push the issue, a jury would probably sympathize with a victim complaining that the ATM offered no way for the coerced user to summon the police. If the bank were to argue that it doesn't provide that capability because it doesn't want to risk a lawsuit from somebody shot by the criminal for attempting to exercise the duty to notify the police implied by the existence of such a feature, the jury would STILL be unsympathetic because at that point, the bank has effectively admitted that to them, the amount withdrawn by the victim at gunpoint is pocket change compared to all possible alternatives. In contrast, there aren't surveillance cameras recording internet purchases. If a cert gets stolen, the instant presumption is that you, the cert's owner, are the one who engaged in fraud, and the burden is on YOU to prove that it was stolen, or your cooperation was coerced, and that you weren't negligent in safeguarding it.
Legislation to enable smart card signatures is nothing new -- I think it's been part of the UCC in the US for almost a decade (or at least, was proposed a decade ago). The problem is, the legislation was so completely lopsided in favor of banks against consumers that you would have had to be financially suicidal and have an economic deathwish to voluntarily participate in it. Even the banks were slightly embarrassed by it, and recognized that it was dead on arrival because no sane consumer would have ever agreed to it.
IMHO, Esperanza (Argentine Antarctica) is just about the most wet-dream-ideal place you could possibly build a datacenter, if you're willing to sink lots of capital into an investment with a 10-20 year payoff horizon. Let's see:
* Fiber-accessibility: not sure whether there's fiber there today (there probably IS), but the area's not glaciated, and the surrounding ocean isn't frozen (lots of icebergs, but the water itself remains liquid year-round), so laying fiber from Esperanza to Tierra del Fuego (or even along the continental shelf all the way to Miami) wouldn't be any big deal if there were actually a reason to do it.
* Climate: 13 degrees Fahrenheit in the winter, 32 degrees Fahrenheit in the summer. Relatively arid & dry year-round. As a practical matter, a winter day in northern Antarctica is nicer than a winter day in Seattle or London. It's just about the perfect climate for servers... cold enough to keep them cool & comfy, but not so bitterly cold that waste heat from the servers can't warm a well-insulated building enough to keep it comfortable for the humans who work there. It's far enough north to always have a few hours of real daylight, and has really, REALLY long days for most of the year... basically 24 hours in the summer.
* Population: OK, it's not a big city (or even a small town) by any means... but it has one big advantage (for aspiring corporations) over comparable settlements run by the United States, Britain, and other countries: Argentina considers it to be a normal city sitting on land that's just another part of Argentina, as opposed to a holy, sacred (if chilly) ecological garden of eden that's off-limits to commercial development (it is, in fact, in Antarctica). I'm sure that if Microsoft showed up with a few billion dollars and a Chinese construction company or two, Argentina would eagerly welcome them with open arms.
* Government: not the United States. Once upon a time, that would have been a major drawback. Now, it's almost a selling point.
* Accessibility: in case you overlooked it, no glaciers, so it's accessible by boat (and probably plane) year-round. Overnight air shipment is a logistical possibility.
The only thing it needs is cheap electricity... and in this case, the fact that it's not America might come in handy, because it potentially means fewer insurmountable regulatory obstacles to building a small nuclear power plant (getting permission to do it in the US, or on de-facto US territory in Antarctica, would be politically impossible).
> Can the average PC user (not necessarily the more technically inclined users here on Slashdot) be > trusted not to screw anything up inside a desktop or laptop PC when installing RAM sticks?
Trusted to install? Sure. Trusted to buy the right ones, especially when trying to max out a motherboard's capacity? Er.... maybe. Maybe.
Half the problem is that most motherboard manuals suck at telling owners what differentiates compatible ram from incompatible ram. Sure, they'll list a dozen "certified" Dimm modules that either haven't been available for sale in North America since 2007, cost 16 times as much as anything you could buy from Newegg, or just plain don't even exist in Google's universe, but it's rare for them to come out and say, "if you buy a N-gig DIMM, it has to use low-density chips organized as @{x, y, z}, which means a double-sided module with at least 16 chips". I thought it was a problem that died with PC100 ram, but it appears an entire generation of DDR2 motherboards (thanks to either nVidia and/or Intel, not sure... maybe both, but it sounds more like something Intel would do in the name of value engineering to shave 2c off the price of a $40 board...) impose limits on what DIMM configurations will work that made the mess circa PC100-vs-PC133 look tame and straightforward.
Case in point: my dad's old motherboard ("old" = 3 years), which I had the pleasure of upgrading last summer (not fun). It had four slots, but the ONLY configuration in which you could make it use 8 gigs was a pair of double-sided 4-gig modules. Four 2-gig modules? Nope. Had to be a pair of 4-gig modules of specific density, with nothing whatsoever in the other pair of sockets. And the only way I even found THAT out was because something about how the manual was worded struck me as odd, and within 3 seconds of hitting Google I found page after page of furious users who didn't find out that little detail until AFTER they ordered the ram from newegg. Best of all, 99 times out of 100, online stores like Newegg don't even TELL you what the ram's configuration is besides the timing, and most of the time the pictures are explicitly tagged as "representative", so you can't even go by them.
IMHO, the only thing worse than trying to shop for RAM is trying to wade through the complete mess Intel and AMD have both made of their part-numbering schemes, where you now officially can't even be assured that two chips bearing identical SKUs will benchmark within 10% of each other, let alone within a fraction of a percent. 4 years ago, I thought it was bad. 18 months ago, when I did my latest upgrade, I thought it was intolerable. Right now, you'd almost have to put a gun to my head to make me dedicate a month of my life to finding my next CPU if you tried to talk me into upgrading today, because it's borderline impossible for even someone who's been building his own PCs for more than 20 years to make informed choices about the crap Intel and AMD are both shoving out the door right now (I use the term "crap", because all evidence seems to suggest that they're both doing it so they can charge more for less and get away with it by intentionally spreading confusion about what's actually being sold).
Snoozing serves a semi-useful purpose. If you're in the deepest phase of sleep, when the alarm goes off, it might take 30 seconds or more to even become *aware* of its significance, and if you were to get out of bed immediately at that point, you'd be borderline sleepwalking. An effective alarm is one that requires some degree of puzzle-solving skill and physical action to turn off the alarm for good, but increasingly demanding activity with shorter quiet intervals to hit 'snooze'. Ideally, to turn off an alarm "for real", you should have to be physically out of bed and on the other side of the room. By turning the alarm/radio back on every few minutes, it keeps you from drifting back into deep sleep, so each time you're progressively more awake than before. If it annoys you and gets your adrenaline flowing, so much the better.
Also, the most effective strategy for winning things at the lowest price on eBay is sniping. If you place your genuine max bid well ahead of the auction's end, there's a good chance that someone's going to end up outbidding you by 13 cents at the last second. If you wait until there's literally no time for anybody else to react to your bid, you achieve two goals -- nobody else has time to outbid your winning bid, and you yourself can't get caught up in auction fever and raise your bid above your pre-decided max bid either if it ends up losing. I don't think I've *ever* won an auction on eBay by placing a max bid days before it ended and walking away. On the other hand, I win about 85% of the auctions I snipe.
Sorry, Officer, but I was *not* texting. I'm blogging with the Wordpress app on my Android phone, and live-tweeting status updates to my followers. Twitter couldn't geolocate me because my Epic4G's GPS was borked by Samsung & Sprint. I accidentally swerved across the middle lane and almost hit the old lady on a scooter because I had to disable 4G so Sprint's firewall won't block the GPS ephemera download, dial *#1472365# to get the phone in debug mode, forcibly flush the cache, launch the diagnostics, wait for it to get a fix, then exit back, re-enable 4G, and finally launch the twitter client. YOU try doing that with one hand...
> HA, that will be the joker in this game that will conquer it all!!
I know it's meant as a joke... but honestly? It depends on whether the sideloading exploits that enable users to bypass the official market requirements for apps ends up being durably defeated. Microsoft has a long history of winking at carriers, then making developers happy by shipping phones with the equivalent of a bathroom-door lock that can be trivially defeated by anyone who bothers to read Slashdot on a regular basis. In D&D terms, Apple, AT&T, and Verizon are all firmly Lawful Evil. Google is Chaotic Good (think: Team America World Police), and Microsoft & Sprint are arguably Neutral Evil or Chaotic Neutral (I don't remember which one is the category of "wants to be bad, but ends up being accidentally good or mostly harmless anyway").
> lifetime ban +++ they crush the vehicle being driven into a cube and deposit it on the driver's front lawn
Big, huge problem -- most people don't actually own their cars. Banks do. Banks purchase cars on behalf of customers, then allow the customer to use them while they're being paid off. It's called a secured asset, and it's the reason why car loans have interest rates in the neighborhood of 5-8%, and credit cards (unsecured loans) have interest rates in the neighborhood of 29-35% or more.
If banks were to start seeing 1% or more of their cars get seized and crushed, EVERYONE would end up paying higher prices for car loans... probably in the form of higher interest rates, possibly in the form of expensive add-on insurance policies priced according to the outstanding balance on the loan required by banks to protect themselves from losses.
The biggest single problem with DUI laws are their "all or nothing" penalties. I don't think there's anybody who'll seriously challenge the fairness of throwing the book at someone who blows.15, or who blows.10 after being involved in an accident. Especially not if it's happened before. If the government wants to reduce drunk driving (instead of merely scoring political victories against a group that's largely without political power), it could go a LONG way by doing things like:
* Making it illegal, point blank, to tow cars that have been parked for fewer than 12 hours. In Miami Beach, it's common for parking lots to charge exhorbitant amounts of money to park, then turn around and have every car still in the lot at 6:30am towed. It's one thing to make somebody pay normal business-hour parking rates for leaving their car parked in the lot past 6:30am. It's another thing entirely to put thousands of moderately drunk individuals in a position where they're forced to choose between a small, but real, risk of getting caught driving with.09 blood alcohol, and the certainty of getting hit with a $250+ towing bill plus everything that goes along with it. Miami Beach allows it, partly because it takes a hefty chunk of all those towing fees and makes a lot more money by practically forcing moderately drunk individuals to risk driving home than it would by encouraging them to leave their cars overnight, take a cab home, and maybe take advantage of being in South Beach the following afternoon to patronize a restaurant.
* Have a police officer present at major parking lots, and offer the following deal if you agree to a breath test before even getting into the car: if you blow over.08, there WILL be a police officer waiting for you around the next corner if you insist upon getting in your car and driving away anyway. If you blow.06 through.08, you'll have two choices: wait until you manage to get it below.06, or be electronically added to the list of drivers flagged for special attention during the next 6 hours (not permanently, but definitely for the next few hours). The grand prize (and reason for consenting) is for drivers who blow less than.06 -- a timestamped certificate that absolves them of any requirement to consent to a breath or blood test for the next 30 minutes if they encounter a checkpoint within 10 miles, 60 minutes if they encounter a checkpoint 10-20 miles away, and 90 minutes if they encounter a checkpoint more than 20 miles away. Essentially, the driver avoids the legal risk of abnormal metabolization, odd readings, or machines with unknown calibration. He's certified as being less than.60, and that's the end of it for the next 30-90 minutes (the distance-dependent time limit is to deal with the case of someone driving to another location and drinking more after being tested... hardcore alcoholics would still slip through that particular crack, but let's be real... someone who's such a hardcore alcoholic that they'd actively try to game the system and sneak additional alcohol after testing lower is going to get caught sooner rather than later, anyway).
> Of course when you combine that with Midsummer snow in Australia
Midsummer *snow*?!? In *Australia?!? From what I've been told, Australia gets snow in the coldest July winter about as often as Orlando or Rome gets snow in January (not entirely unheard of, but people talk about it for years afterward when it happens). Do you really mean it snowed in the middle of Australia's summer? Or do you mean, "It snowed when it was mid-summer in the northern hemisphere, and the coldest dead of winter in Australia and the rest of the southern hemisphere (where seasons are the opposite of those in the northern hemisphere)?"
I guess this falls into the category of "only on Slashdot...", but yeah, it does seem kind of sad that they didn't go for the gold ring at 32,768. It's obviously not a rounding error, either, because then the news story would have said that there are something like 30000, 32000, 33000, or 35000. 32256 -- so close, yet so far...;-)
127 bottles of of beer on the wall, 127 bottles of beer...
> Most Chinese typists I know use some sort of input means that allows them to type the pinyin (like typing English) and then > select a character corresponding to the pinyin from a list (done with a number at the end, so the hand never leaves the keyboard).
Correction: most nontechnical and older Chinese users of business applications use Pinyin. The Chinese Slashdot crowd uses a much more efficient input method called Wubizixing that breaks characters down into root radicals and strokes. Wubizixing is more or less the Chinese equivalent of touch-typing QWERTY, and someone who's good at it can type the equivalent of 150 English words per minute.
For key-limited scenarios like cell phones, a variant called Wubihua combines stroke-input (like Wubizixing) with character-picking (like Pinyin), and is actually the input method that's the most usable by westerners who can look at a written character and replicate it, but don't actually know how to pronounce the Mandarin or Cantonese word it represents.
> and for centuries before that Latin was the common language of Europe.
It's important to remember that the Latin taught in classrooms would be mostly unintelligible to most people, including Romans, who actually lived in the Roman Empire at its peak. 2,000 years ago, a Roman Senator might have given speeches in Latin, but the language he actually *spoke* with his family and friends would be mostly recognizable today as the dialect of Italian spoken in southern Italy. A wealthy, powerful Roman Senator from Milan might have written equally lofty Latin, but the language HE spoke among family and friends was strikingly similar to the dialect of Italian spoken today in northern Italy. And so on, as you got farther and farther away from Rome. People in what's now modern Spain, France, and Portugal didn't speak Latin... they spoke languages that were quite similar to modern French, Spanish, and Portuguese. They just pretended that "Latin" was the proper and correct way to represent the language they spoke in writing. Latin didn't magically cease to be the world's common language... people just gradually quit pretending that it was the one and only right way to write their own, and came up with alternate grammar and orthography rules that happened to reflect reality better than Latin did.
> I'd be surprised if the next 20 years didn't see a move towards the international language being Chinese (pinyin).
Actually, I think you're 180 degrees off. If anything, a hundred years from now, English will be even more dominant as a language, but will be commonly (if not officially) written with a few hundred Chinese characters used as alternate ways of writing English words.
First, it'll be done to be "cute" or "cool", then gradually start to be done as an alternate way of abbreviating longer English words with ambiguous abbreviations once an entire generation or two has reached college after growing up with Japanese anime and Chinese sci-fi.
I fully expect to see "" (ren) and "" (nu) on bathroom doors across America by 2030. (with lots of confusion among first-time Chinese visitors, since "" (ren) actually means "people", and will probably lead at least a few visitors to think the men's room is around the corner, and the "" (ren) room is for families with small kids).
Pinyin is passe and 20th-century. Even a semi-clueless westerner who can't speak a word of Mandarin or Cantonese to save his life can type Chinese using wubihua and a numeric keypad after a 10-20 minute lesson on stroke order. In fact, Wubihua has a huge advantage over pinyin -- it's the one input method a westerner can USE if he knows what a character looks like, but doesn't know how to actually say the word it represents.
Now, if only Slashdot would allow Chinese to actually be included in postings without getting mangled...
> Uhm Hindi [wikimedia.org] is the official language of India.
Roughly half of my coworkers were born (and grew up) in India. I think *one* of them might actually be capable of sitting down at a computer running Windows and sending email typed in Hindi. My boss told us point blank that almost nobody in India actually types Hindi, because it's too much of a pain relative to typing English. That's not to say Indian computer users wouldn't *prefer* to have things like menus and text DISPLAYED in Hindi, but as a practical matter, it's almost unheard of for a computer user in India to go to the trouble of switching to Hindi input for something like email, because everyone he or she is likely to care about probably knows English better than the dialect of Hindi the writer would use *anyway*
> Admins upon security advice upload settings which make the device unusable. In that case "reporting compliance" > while it is not from the user viewpoint is actually a useful feature.
There's actually a useful compromise that's so obvious, it completely blows my mind that it appears to have not even occurred to Microsoft -- keep the corporate data on the server, and give the end users Android and iPhone customized RDP clients that connect to a hosted email app on the server (with the ability to launch intents to alert users to new email when the app isn't actively unlocked, without actually passing sensitive data to the phone itself... ie, "New email" (or maybe "New email from ${name}", if enabled by server policy). Users want to read email? Strong authentication. Phone inactive in "mail" mode for too long? Hide the mail app's view (requiring authentication to re-activate), but leave the rest of the damn phone alone.
This isn't rocket science. Is it bulletproof? Of course not. Any rooted phone can do screenshots. But then again, all you'd need to take an effective screenshot of the most locked-down Android/iPhone ever made is an exotic piece of gear called... (drumroll)... a flatbed scanner. Or a DSLR with macro lens. Or a piece of paper and a pen or pencil. Or a voice recorder. The point is, by keeping the data itself on the server, and using the end user's device only to render the current view, you massively reduce the attack surface... and you get to do it without making the end user completely miserable.
Slashdot Mirror
> It would be IE vs Netscape all over again
Except this time, Netscape is metaphorically Microsoft's product, and IE smells like fresh, hot gingerbread. Mmmmm!
Take away Google, and the streets will riot, Take away Bing, and... er... um... someone might eventually notice. Maybe.
Hotmail? Is that actually used for anything besides MSDN SSO credentials anymore?
IE? (rolling on the floor, gasping for breath amidst near-lethal amounts of laughter). People with *Windows Mobile* didn't even use Internet Explorer before Android came out. We used Opera, and paid for it, because Pocket IE sucked like a whore with braces. I'm sure Microsoft did a much better job with IE on WP7... but then again, a chimp with a Commodore 64 and a pirated copy of GEOS could probably improve upon PIE in its WM6.x incarnation.
MS Office Suite? Meh. Apps to view and edit word/excel docs are free and abundant. Outlook still can't do adhoc aliases properly, and Android can be induced to lie about its authentication capabilities so you won't have to indulge your company's IT department's wet dreams by entering a 16-digit passcode before it allows you to answer an incoming phone call.
What Microsoft REALLY needs to do a better job of is tracking mouse movement (specifically, acceleration, ballistics, etc) to do a better job of discerning intent when you go to grab something... and a better job of adaptively figuring out over time whether its assumptions about your intent are right or wrong. It really seems like every new version of WIndows leaves me fighting and frustrated with it a tiny bit more.
Here's a real-world concrete example. Suppose the mouse pointer is approaching the right edge of a window that's maximized to the left panel of a multi-monitor setup. The mouse pointer slows down, and seems to also be approaching the scrollbar. The left button gets pressed, and the mouse moves in a direction that's mostly upwards. Well, except the pointer overshot the edge a bit, and the left click technically occurred 2 pixels into the window on the monitor adjacent to the right. Taken in isolation, Windows has no real choice but to assume the user meant to click the pixel on the other monitor even though it contextually makes no sense. But combined with the observed ballistics (slowing down, slight arc towards the scrollbar, motion after left-click that makes more sense as a scroll-gesture than a... well... meaningless gesture), it's obvious what the user meant to click. And for the most part, Windows, seems to be completely oblivious to it.
Now, for a counter-example: trying to select text without adjacent whitespace. For me, Windows (Word and Outlook in particular) NEVER seems to get this right. I'll click at the right starting point, letting go and starting over if I'm not happy with it. Then I'll start highlighting. But way too often, it'll stop selecting a character or two short of where I want. If I keep moving the mouse, it'll grudgingly select the remaining characters... but feels compelled to ALSO ignore my hard initial-selection work and expand the other end of the selection too. Dammit. I'll then spend the next 10 seconds fighting with it trying to select the text I REALLY want. Half the time, I'm forced to give up, let it select the damn adjacent whitespace, and edit it away after I paste. It annoys me to no end.
In the end, it feels like Windows has simultaneously gotten worse in two directions. It forces its opinion on me without learning from its mistakes or giving me the option to beat it into submission so it quits interfering, and simultaneously forces selection with almost single-pixel precision to make increasingly-dense window gadgets work. I'll admit that Java is even worse in this regard, and Linux (or at least Gnome/Compiz) doesn't seem to be any better, but it's still annoying as hell.
You know, if I could go back in time and try to change somebody's mind, I'd make it my mission to try and convince the French to define the meter such that 24mm were precisely equal to one imperial inch. Why? Because then, instead of having two sets of tools requiring two entirely different sets of screws and bolts, we'd only need one, regardless of whether you prefer metric or imperial, because measurements in either system would have precise, nice, exact equivalents in the other. 1 inch? 24mm. 3/4 inch? 18mm. 1/2 inch? 12mm. 1/4 inch? 6mm. 5/8 inch? 15mm. 7/16 inch? OK, not as pretty, but still a reasonable 10.5mm. 23/32 inch? Ugly, but still-tolerable 17.25mm (compared to an abomination like 18.25625mm). You'd basically have to go to 64ths of an inch before you ever had to deal with fractional millimeters besides .25mm, .5mm, or .75mm.
OK, so it wouldn't have been 1/10000000th the distance from the equator to the poles... but it turns out, they were wrong about the distance, so the meter's length ended up being as arbitrary as the length of some ancient king's foot anyway.
Check eBay, it's littered with sub-600MHz Android tablets with barely enough flash to hold 1.6, barely enough ram to boot, no Android Market, and likely to spend most of their lives as... digital picture frames, because most users aren't masochistic enough to suffer with them for more than a few days once the novelty wears off.
Don't write them off, though. They might be useless as general-purpose Android devices, but they're cheap enough to use as single-purpose devices (home theater remotes, family photo albums, home automation controllers, interactive cookbooks, etc) whose ultimate use just happens to have not been carved in stone (etched in silicon?) at the factory.
Nevertheless, if you really want a general-purpose Android tablet to play with NOW, don't touch ANYTHING that doesn't ship with 2.1 or better, have a 1GHz or faster CPU, enough battery life to actively use for 3-4 hours, and still have enough of a charge after laying ("off") on a table or sofa cushion for a day or two to use for 5-10 minutes before you HAVE to put it back on the charger. Try to find one that has 1280x800 or better resolution, because that's the magic point where you can almost read two O'Reilly-sized pages side by side from a pdf file. Not coincidentally, demanding better than 480x800 ALSO happens to weed out most of the tablets that are too slow/limited to be worth bothering with right now.
Oh, and just to warn everyone... don't buy a Galaxy Tab unless you're absolutely delighted with it as it exists RIGHT NOW, because Samsung has already fucked millions of customers who made the mistake of buying a Captivate, Vibrant, Epic4G, Fascinate, or Mesmerize. Maybe it's Samsung's fault, maybe it's the carriers' fault, but either way, if you buy a Galaxy Tab, it's coming from Sprint, Verizon, AT&T, Verizon, or some regional carrier. If they don't even care enough about their millions of PHONE customers to upgrade them to Froyo and fix the goddamn dysfunctional GPS, does anyone really think they're going to pay more attention to a few hundred thousand tablet purchasers?
> Mobile Windows didn't have such glaring problems with malware stealing from the user.
That's mostly because statistically, there weren't enough Windows Mobile users (or PalmOS users, or Symbian users, for that matter) to be worth the time of organized crime.
The problem with Android isn't the fact that the source is available to peruse, it's the fact that manufacturers and American carriers do their best to make upgrades as difficult as possible despite Android's open-source Linux roots. An exploit like this barely gets a yawn from Nexus One users, because someone will update it before any real exploits based on this ever become a problem. In contrast, owners of American Samsung Galaxy S phones will be shitting bricks, because we're still waiting for a fucking kernel that works with Froyo. Or at least leaked CDMA loadable kernel modules compatible with a 2.6.32 kernel so we can build our own without losing basically all the hardware drivers it needs to work properly.
Sidetrip: Unlike Windows, Linux makes no effort to maintain a stable ABI between versions. Simplified a bit, this basically means that a loadable kernel module (the Linux analog to a hardware driver) that's built for a 2.6.29 kernel will probably crash and burn on a 2.6.32 kernel. The official Linux party line is that it makes it harder for manufacturers to keep drivers proprietary, and motivates vendors to release source for their drivers so it can be automatically rebuilt for each new kernel release. The cold American consumer reality is that the Android Emperor is nude. The Nexus S can't do 4G on T-Mobile, is fundamentally incompatible with Sprint and Verizon, and AT&T's slow, capped, expensive 3G isn't even a real option. We're stuck with an allegedly-open operating system inextricably bound to hardware that's more locked down and proprietary than an iPhone, and all we can really do is hope some of Linux's core developers also own Android phones and are starting to really, really feel some of the ABI pain themselves on a daily basis.
Put another way, here's a more technical summary of the problem:
* Samsung has released source to its kernel and loadable kernel module drivers, but the LKM source won't build against any known 2.6.32 kernel due to missing dependencies.
* The .ko modules themselves were built against the ABI of a specific build of 2.6.29 that changed enough with 2.6.32 for most of them to crash and burn if you try using them with a 2.6.32 kernel.
* Froyo and Gingerbread have dependencies on the 2.6.32 kernel. You can cobble together a FrankenBuild that sort of works with a 2.6.29 kernel, but it'll never be a True Froyo/Gingerbread, and will always have bugs hidden below the surface veneer.
Metaphorically, an American Samsung Galaxy S trying to run Froyo is kind of like a laptop that shipped with Windows 98 and a winmodem. The unfortunate user upgrades it to XP himself, then discovers that the winmodem only has drivers for Win98. Through some miracle, the winmodem drivers have their "source" released, but that source requires a thirdparty library called LunexantProprietaryLib that isn't included, and won't build without it. After lots of hacking, the user manages to cobble together drivers that will allow the modem to limp along at 9600 baud by pretending it's an older version of the chipset, but getting it to do 56k without official drivers is hopeless. And if, by some miracle of god, a never-released copy of drivers for XP get leaked despite the determination of the manufacturer to keep it unavailable through the perverse logic that fucking their customers will somehow encourage them to buy a newer model from the same company that screwed them less than a year earlier (instead of buying one made by just about ANYBODY else), the user discovers that the drivers needed for 3D acceleration have the same problem as the Winmodem, and it's back to square one.
What Google really needs to do is define an ABI thunking layer and require that any and all device drivers
Not to mention, passengers in a train have self-help options in the event of catastrophe that passengers in a plane lack. They might or might not be useful, but as a practical matter, if you're on a 747 that gets blown in half 3 miles up, and by virtue of being in the rear tail section that gets blown away, well... you'll get to have 3-5 terrifying additional minutes of life that your fellow passengers didn't. Ultimately, though, you're as screwed as the rest of them. In contrast, if you're a passenger on a train that gets blown in half and your part skids away from the rest... well, you might actually live to tell about it.
There's also the fact that American passenger trains are slightly unique. By law, they have to be rolling bank vaults capable of surviving a head-on collision with a mile-long freight train. 99.9% of the time, that's stupid and bad, but when things like explosives get involved, it *does* tend to come in handy.
The main problem with recent (past year or two) HSR proposals is that they take a generally good thing, then ruin it by demanding an expensive level of immediate perfection that will ultimately keep it from ever really succeeding by virtue of both cost and limited scope. Take Florida. By all means, build the tracks between Orlando and Tampa to 220mph HSR standards. They're going to be around for a really long time, and it's worthwhile to do that particular job right the first time around. It won't raise the cost by much, and it'll seriously reduce the construction misery 25 years from now when the higher-grade of trackwork is *really* needed. On the other hand, fuck the electrification and trainsets that do 180mph, but can ONLY run on their dedicated, purpose-built tracks like a big amusement park ride. By all means, design the tracks to be electrified *later*, but for now, stick with diesel. You *don't* electrify a rail line that has one train per hour in each direction. It's economically insane. Not even the former Soviet Union was *that* crazy (at least, on a large scale; I'm sure there were a few tracks here and there that were electrified for the sake of politics, but I'm pretty sure the Trans-Siberian Railroad STILL isn't fully-electrified all the way to the Pacific). Instead, connect the new track (which HAS to be built for any kind of meaningful passenger service to Tampa from either Orlando or Miami, because the last 20 miles of existing track into Tampa ARE very, very intensively used by CSX for freight) to the existing CSX tracks from Auburndale down to West Palm Beach (which are almost commercially useless for freight), build a second, new track alongside it, and use the cash that WOULD have gone to run semi-useless 180mph trans back and forth between Tampa and Orlando to build a useful passenger rail network with 100-110mph average speeds between Miami, Orlando, and Tampa... maybe even Jacksonville a year or five later. FDOT has studies dating back to before most of us were born that have credibly concluded that 80-110mph passenger service in Florida wouldn't just pay for itself... it would make a real, honest to god outright *profit*. Those same studies invariably conclude that "real" HSR would hemorrhage money forever, mostly because the interest on the construction bonds used to finance the staggering up-front construction costs would crush any hope of it ever achieving financial viability. Florida's politicians need to realize... trainsets are a 20-30 year investment, but trackbed and right of way are forever. It's easy to electrify an existing rail line 25 years later. It's damn hard and expensive to try and straighten out curves after the fact.
Put another way:
* No-compromise "True" HSR between only Orlando and Tampa: crushingly expensive and minimally-useful.
* No-compromise "true" HSR between Miami, Orlando, and Tampa, built NOW: very, very nice... but it would bankrupt FDOT, and would more or less forever write off service to Jacksonville and southwest Florida, let alone Tallahassee (Tallahassee's problematic, just because it's so freakin'
> It has a presence, yes... but doing "just fine"? The iPhone and Androids each have more units in the channel than WP7 has in-channel and
> activated *combined*. This is in spite of the fact that WinMo (in various incarnations) have been for sale for (almost) a decade.
They don't exist on Sprint and Verizon, yet. In the United States, that makes them commercially irrelevant because T-Mobile is tiny, and NOBODY voluntarily uses AT&T unless they're shackled to a pre-Verizon iPhone.
Likewise, Microsoft's success or failure with Windows Mobile is largely irrelevant. I'm sure there are 3 or 4 Windows Mobile users left who haven't jumped to Android or IOS, but I don't know them personally. The fact that Windows Phone is metaphorically "Sidekick5dotnet" and doesn't run Windows Mobile software makes almost any link between them irrelevant as well.
Still, as others have noticed, Microsoft can afford to throw monkey wrenches into the plans of others for a really, really long time. In the long run, Microsoft probably WILL recover a chunk of its marketshare with specific respect to the enterprise market if they ever get their act together and make Windows Phone integrate *seamlessly* with Exchange Server and Microsoft's management infrastructure.
There's one specific case where it's easy to replicate a visible effect: GSM noise. It drives air traffic controllers insane, because they get hammered by it all day. The irony? It's only really a problem when the GSM radio lies within a few feet of the microphone. Yup, that's right folks... basically all of the GSM noise that assaults ATC daily comes from the (powered up) Blackberries of PILOTS in the cabin. Even then, the solution is easy: use Sprint or Verizon phones, because CDMA doesn't cause the problem (technically, a GSM phone operating exclusively in UMTS mode wouldn't, either... but I don't think it's actually possible to selectively disable only legacy TDMA-based GSM on a Blackberry or iPhone and force it to use ONLY WCDMA-based UMTS).
The ban is almost entirely a matter of regulatory inertia and risk of lawsuits. Since the bans are universal, no airline wants to risk paying higher damages in a lawsuit if there's a crash and someone is able to convince a jury that their policy of allowing electronic devices theoretically increased their liability by even .000001%. On the other hand, if airlines could install picocells that made it impossible to connect to a carrier's towers, but enabled them to collect $1/minute roaming charges, you'd see any hint of a ban eliminated within a matter of days.
They're probably generating a 22KHz sine wave and running it through a transformer to boost the voltage from ~0.7 volts (more or less what you'd get if you drove the output DAC at the highest power output it could sustain with a non-fully-charged battery without clipping) to 3.3 or 5.0 volts. Yeah, you can do the same thing with about $10 worth of active components, but if you can easily generate a real sine wave to start with, you can basically replace the solid-state charge pump with a transformer, two diodes, and a capacitor.
As far as parasitic power goes (a fancy way of saying, "Drawing power from the data bus"), one way I know of is to drive your data line at two voltages... say, 6v or less for 0, 9v or higher for 1. At the other end, use a linear voltage regulator to convert it back down to 5.0v for power, and connect it through a resistor to an optoisolator whose other end is sampled by a GPIO pin. In real life, it's a little more complicated, because the capacitors you need would affect the signal, but I'm pretty sure the general idea would work (at least, for slow data rates).
My favorite back-channel approach to overloading power lines for data (in scenarios where you aren't trying to cover distances where voltage drops are an issue) is to add a diode bridge, and sample one output leg of it with an i/o pin. For example, suppose you have a microcontroller that has two pins: one for 5v DC, and one for ground. Now, suppose you need a way to convey a single bit of information to that chip in situations that really only matter during development or debugging. By sensing the output of one of the diode bridge's leg, you can tell whether the user connected the power normally or backwards, and act accordingly.
My own experiments have been Android-related (exploring the headphone jack as a gamepad input), but I came up with two obvious configurations for using a jack with ground, mic (input), left audio, and right audio:
* SPI (.7v logic): left audio = clock; right audio = master out, slave in (phone is master, obviously); mic = master in, slave out. Ground=ground. ~8-14 months from now, my Epic4G is going to be my Bioloid's new backpack-mounted controller :-)
* UART. left audio = .7v sine wave for power, right audio = TxD, mic = RxD, ground = ground. Obviously we're talking about nonstandard voltage levels and not RS232.
Other possibilities:
power circuit using left audio sinewave, use right audio to modulate infrared LED and make a cheap remote control.
bitbang NTSC s-video chroma & luma signals via right and left audio connectors... or mono audio + composite video. If a 20MHz 8-bit CPU can pull off color video better than most late-1970s videogames, an Android phone with 1GHz CPU hijacked into single-user Linux mode devoting its full attention to bitbanging the i/o should be able to pull off graphics at least as well as the Atari 2600. OK, pretty useless, but damn it would make a cool video on Youtube ;-)
If the left audio, right audio, and mic pins are really GPIO with data direction registers somewhere, you could hack a decent 3-channel logic analyzer (or 2 channels + trigger). If the phone has a crossbar switch between the USB port and the rest of the phone, you might even be able to hijack the two pins normally used for USB data and get two more inputs. Add a bit more hardware, like a FTDI usb bridge chip with FIFO and bitbang-mode capabilities, and an old Android phone (or iPhone) capable of activating USB-OTG would make a totally kick-ass pocket logic analyzer.
Argh. Accidentally deleted an entire section and didn't notice until it was too late.
As far as how incoming calls are handled when there's an active data session, it depends entirely on the phone. The CDMA2000 standard itself has a perfectly good way for phones to poll for incoming calls, text messages, and voicemail notifications when there's an active EVDO data session. The problem is that historically, most phones get it wrong and totally fuck up the implementation.
Example: Sprint PPC-6700 and its evil twin, the Verizon xv6700. The original radio firmware took a long time to switch between EVDO and CDMA voice, so Microsoft (or whomever was customizing Windows Mobile 5 for the phone) decided to handle incoming voice calls during EVDO data sessions by pre-emptively breaking the data connection, then lamely asking (too late) whether you wanted to end the data session to take the call. Angry users eventually figured out how to hack the phone to make it refrain from breaking the connection until after you told it whether you wanted to take the call. Unfortunately, that hack caused lots of calls to end up going to voicemail at the last second, because the phone couldn't switch to CDMA voice mode in time to take the call. You'd get the alert, acknowledge that you wanted to take the call, and have it go to voicemail anway. The arms race continued, and we figured out how to make the phone poll the tower for incoming calls more frequently. It only increased the time we had to accept the call by about a half second... but that half second was enough to keep the majority of calls from falling through the crack and going to voicemail.
Enter the Sprint Hero. In its virgin 1.5 glory, it didn't even bother to ask whether you wanted to take the incoming call... if you had an active data connection, the call would silently go straight to voicemail. Worse, you wouldn't even find out about it until you laid off the data use long enough for the phone to close the EVDO session and go back into 1xRTT mode. After lots of hatred from angry customers, Sprint made it less bad sometime around Thanksgiving so calls still went straight to voicemail, but at least the phone properly polled for new text messages and voicemail notifications during active EVDO data sessions. As far as I know, Sprint's official ROMs *never* fixed the problem entirely, but I quit caring the day I discovered AOSP anyway (which did, in fact, fix the problem).
The Epic4G mostly seems to handle the situation properly... if you get an incoming voice call while you have an active EVDO session, it notifies you immediately, but it doesn't break the data connection unless you actually take the call. And if your data session is via wifi or wimax, it continues to work.
Truth be told, I'll be shocked if the iPhone DOESN'T handle this situation gracefully. Where things get sticky is how it will handle the opposite scenario (user tries to make outgoing phone call while running one or more networked apps). I suspect American apps will handle it better than European apps, just because EDGE and GPRS are facts of life in the US, but you can allegedly go kayaking through the fjords in northern Scandinavia, take pictures of a real reindeer, and share them with your friends 8,000 miles away via UMTS, even if the cell site HAS to be solar-powered because the nearest power line is 100km away ;-)
Here's the deal: officially, CDMA2000-EVDO can't simultaneously do voice and data. Specifically, it's impossible to simultaneously have an EV-DO data session and circuit-switched CDMA voice call using the equipment Verizon and Sprint purchased from Qualcomm. However, the line has gotten a little blurred lately, especially with rooted Android phones (and almost certainly, jailbroken CDMA iPhones by summer). Five years ago, the answer to "Can you do simultaneous voice and data with EVDO" was unambiguously "No, it's impossible." Now, it's more like, "well... it depends. Do you have a rooted/jailbroken phone? Are you OK with spending another $10-25/month for thirdparty VoIP service? Good. In that case, the answer is, "Yes, kind of."
There are a few projects to enable users with thirdparty VoIP service and rooted Android phones to use it with their phones. Basically, here's how it works:
* You sign up for VoIP voice service, get a phone number, and configure it to forward incoming calls to your cell phone's number.
* You run an app on your phone that intercepts "incoming call" notifications when there's an active EVDO data session.
* When the app sees an incoming call, it checks to see whether it's from the VoIP number. If it is, it quickly establishes its own VoIP session over the EVDO data connection, launches the VoIP phone app, and lets you take the call without losing your data session.
The catch is, the apps (AFAIK) can't do traffic-shaping on your local network connection, so it's up to you to make sure that nothing else running on the phone saturates it and causes it to drop too many UDP packets. The last I checked, they were still kind of raw, unpolished, and required Android phones that were rooted to really work reliably (or at all)... but I don't see any obvious insurmountable reason why they couldn't work just as well with a jailbroken Verizon iPhone.
> If someone can sign your name on a paper and send it by mail you'd be fucked to. ...
Actually, no. You could legitimately argue (in court, if necessary) that your signature was forged. Forgery is so common, assertions of it in court are almost automatically accepted by juries as credible unless the party claiming it's legitimate can bend over backwards and demonstrate (through supporting evidence, like driver's license data, video surveillance footage showing the individual perform the transaction, etc) overwhelming evidence that it's legitimate.
Smart card-based certificates upset that delicate balance of power. They don't prove that it was signed by you, but they do prove (almost beyond doubt) that something was signed by someone with physical possession of your card/cert and knowledge of its security code. Thus, they instantly shift the issue from claims by the victim that his signature was forged (something that's happened throughout human history, is commonplace, and an easy defense for consumers to successfully raise in court) to claims by the banks that you were negligent in your handling of the certificate and/or its security code. As a consumer, you have basically no duty to prevent someone else from forging your signature, because you can't. And the scenarios where banks could claim you were negligent would be almost impossible for them to prove. In contrast, with the cert/card, if anything goes wrong, banks have a MUCH easier time of shifting liability to you, the consumer.
You could argue that a similar situation exists with ATM cards, but ATMs have an advantage (for consumers) that internet transactions don't -- pervasive video surveillance. If a criminal coerces you to give up your PIN code, it's likely to be pretty easy to prove his involvement and demonstrate coercion. If the criminal is out of view, but the victim claims otherwise, the bank's in an awkward position. If the bank were to push the issue, a jury would probably sympathize with a victim complaining that the ATM offered no way for the coerced user to summon the police. If the bank were to argue that it doesn't provide that capability because it doesn't want to risk a lawsuit from somebody shot by the criminal for attempting to exercise the duty to notify the police implied by the existence of such a feature, the jury would STILL be unsympathetic because at that point, the bank has effectively admitted that to them, the amount withdrawn by the victim at gunpoint is pocket change compared to all possible alternatives. In contrast, there aren't surveillance cameras recording internet purchases. If a cert gets stolen, the instant presumption is that you, the cert's owner, are the one who engaged in fraud, and the burden is on YOU to prove that it was stolen, or your cooperation was coerced, and that you weren't negligent in safeguarding it.
Legislation to enable smart card signatures is nothing new -- I think it's been part of the UCC in the US for almost a decade (or at least, was proposed a decade ago). The problem is, the legislation was so completely lopsided in favor of banks against consumers that you would have had to be financially suicidal and have an economic deathwish to voluntarily participate in it. Even the banks were slightly embarrassed by it, and recognized that it was dead on arrival because no sane consumer would have ever agreed to it.
IMHO, Esperanza (Argentine Antarctica) is just about the most wet-dream-ideal place you could possibly build a datacenter, if you're willing to sink lots of capital into an investment with a 10-20 year payoff horizon. Let's see:
* Fiber-accessibility: not sure whether there's fiber there today (there probably IS), but the area's not glaciated, and the surrounding ocean isn't frozen (lots of icebergs, but the water itself remains liquid year-round), so laying fiber from Esperanza to Tierra del Fuego (or even along the continental shelf all the way to Miami) wouldn't be any big deal if there were actually a reason to do it.
* Climate: 13 degrees Fahrenheit in the winter, 32 degrees Fahrenheit in the summer. Relatively arid & dry year-round. As a practical matter, a winter day in northern Antarctica is nicer than a winter day in Seattle or London. It's just about the perfect climate for servers... cold enough to keep them cool & comfy, but not so bitterly cold that waste heat from the servers can't warm a well-insulated building enough to keep it comfortable for the humans who work there. It's far enough north to always have a few hours of real daylight, and has really, REALLY long days for most of the year... basically 24 hours in the summer.
* Population: OK, it's not a big city (or even a small town) by any means... but it has one big advantage (for aspiring corporations) over comparable settlements run by the United States, Britain, and other countries: Argentina considers it to be a normal city sitting on land that's just another part of Argentina, as opposed to a holy, sacred (if chilly) ecological garden of eden that's off-limits to commercial development (it is, in fact, in Antarctica). I'm sure that if Microsoft showed up with a few billion dollars and a Chinese construction company or two, Argentina would eagerly welcome them with open arms.
* Government: not the United States. Once upon a time, that would have been a major drawback. Now, it's almost a selling point.
* Accessibility: in case you overlooked it, no glaciers, so it's accessible by boat (and probably plane) year-round. Overnight air shipment is a logistical possibility.
The only thing it needs is cheap electricity... and in this case, the fact that it's not America might come in handy, because it potentially means fewer insurmountable regulatory obstacles to building a small nuclear power plant (getting permission to do it in the US, or on de-facto US territory in Antarctica, would be politically impossible).
> Can the average PC user (not necessarily the more technically inclined users here on Slashdot) be
> trusted not to screw anything up inside a desktop or laptop PC when installing RAM sticks?
Trusted to install? Sure. Trusted to buy the right ones, especially when trying to max out a motherboard's capacity? Er.... maybe. Maybe.
Half the problem is that most motherboard manuals suck at telling owners what differentiates compatible ram from incompatible ram. Sure, they'll list a dozen "certified" Dimm modules that either haven't been available for sale in North America since 2007, cost 16 times as much as anything you could buy from Newegg, or just plain don't even exist in Google's universe, but it's rare for them to come out and say, "if you buy a N-gig DIMM, it has to use low-density chips organized as @{x, y, z}, which means a double-sided module with at least 16 chips". I thought it was a problem that died with PC100 ram, but it appears an entire generation of DDR2 motherboards (thanks to either nVidia and/or Intel, not sure... maybe both, but it sounds more like something Intel would do in the name of value engineering to shave 2c off the price of a $40 board...) impose limits on what DIMM configurations will work that made the mess circa PC100-vs-PC133 look tame and straightforward.
Case in point: my dad's old motherboard ("old" = 3 years), which I had the pleasure of upgrading last summer (not fun). It had four slots, but the ONLY configuration in which you could make it use 8 gigs was a pair of double-sided 4-gig modules. Four 2-gig modules? Nope. Had to be a pair of 4-gig modules of specific density, with nothing whatsoever in the other pair of sockets. And the only way I even found THAT out was because something about how the manual was worded struck me as odd, and within 3 seconds of hitting Google I found page after page of furious users who didn't find out that little detail until AFTER they ordered the ram from newegg. Best of all, 99 times out of 100, online stores like Newegg don't even TELL you what the ram's configuration is besides the timing, and most of the time the pictures are explicitly tagged as "representative", so you can't even go by them.
IMHO, the only thing worse than trying to shop for RAM is trying to wade through the complete mess Intel and AMD have both made of their part-numbering schemes, where you now officially can't even be assured that two chips bearing identical SKUs will benchmark within 10% of each other, let alone within a fraction of a percent. 4 years ago, I thought it was bad. 18 months ago, when I did my latest upgrade, I thought it was intolerable. Right now, you'd almost have to put a gun to my head to make me dedicate a month of my life to finding my next CPU if you tried to talk me into upgrading today, because it's borderline impossible for even someone who's been building his own PCs for more than 20 years to make informed choices about the crap Intel and AMD are both shoving out the door right now (I use the term "crap", because all evidence seems to suggest that they're both doing it so they can charge more for less and get away with it by intentionally spreading confusion about what's actually being sold).
Snoozing serves a semi-useful purpose. If you're in the deepest phase of sleep, when the alarm goes off, it might take 30 seconds or more to even become *aware* of its significance, and if you were to get out of bed immediately at that point, you'd be borderline sleepwalking. An effective alarm is one that requires some degree of puzzle-solving skill and physical action to turn off the alarm for good, but increasingly demanding activity with shorter quiet intervals to hit 'snooze'. Ideally, to turn off an alarm "for real", you should have to be physically out of bed and on the other side of the room. By turning the alarm/radio back on every few minutes, it keeps you from drifting back into deep sleep, so each time you're progressively more awake than before. If it annoys you and gets your adrenaline flowing, so much the better.
Also, the most effective strategy for winning things at the lowest price on eBay is sniping. If you place your genuine max bid well ahead of the auction's end, there's a good chance that someone's going to end up outbidding you by 13 cents at the last second. If you wait until there's literally no time for anybody else to react to your bid, you achieve two goals -- nobody else has time to outbid your winning bid, and you yourself can't get caught up in auction fever and raise your bid above your pre-decided max bid either if it ends up losing. I don't think I've *ever* won an auction on eBay by placing a max bid days before it ended and walking away. On the other hand, I win about 85% of the auctions I snipe.
Sorry, Officer, but I was *not* texting. I'm blogging with the Wordpress app on my Android phone, and live-tweeting status updates to my followers. Twitter couldn't geolocate me because my Epic4G's GPS was borked by Samsung & Sprint. I accidentally swerved across the middle lane and almost hit the old lady on a scooter because I had to disable 4G so Sprint's firewall won't block the GPS ephemera download, dial *#1472365# to get the phone in debug mode, forcibly flush the cache, launch the diagnostics, wait for it to get a fix, then exit back, re-enable 4G, and finally launch the twitter client. YOU try doing that with one hand...
> HA, that will be the joker in this game that will conquer it all!!
I know it's meant as a joke... but honestly? It depends on whether the sideloading exploits that enable users to bypass the official market requirements for apps ends up being durably defeated. Microsoft has a long history of winking at carriers, then making developers happy by shipping phones with the equivalent of a bathroom-door lock that can be trivially defeated by anyone who bothers to read Slashdot on a regular basis. In D&D terms, Apple, AT&T, and Verizon are all firmly Lawful Evil. Google is Chaotic Good (think: Team America World Police), and Microsoft & Sprint are arguably Neutral Evil or Chaotic Neutral (I don't remember which one is the category of "wants to be bad, but ends up being accidentally good or mostly harmless anyway").
> lifetime ban +++ they crush the vehicle being driven into a cube and deposit it on the driver's front lawn
Big, huge problem -- most people don't actually own their cars. Banks do. Banks purchase cars on behalf of customers, then allow the customer to use them while they're being paid off. It's called a secured asset, and it's the reason why car loans have interest rates in the neighborhood of 5-8%, and credit cards (unsecured loans) have interest rates in the neighborhood of 29-35% or more.
If banks were to start seeing 1% or more of their cars get seized and crushed, EVERYONE would end up paying higher prices for car loans... probably in the form of higher interest rates, possibly in the form of expensive add-on insurance policies priced according to the outstanding balance on the loan required by banks to protect themselves from losses.
The biggest single problem with DUI laws are their "all or nothing" penalties. I don't think there's anybody who'll seriously challenge the fairness of throwing the book at someone who blows .15, or who blows .10 after being involved in an accident. Especially not if it's happened before. If the government wants to reduce drunk driving (instead of merely scoring political victories against a group that's largely without political power), it could go a LONG way by doing things like:
* Making it illegal, point blank, to tow cars that have been parked for fewer than 12 hours. In Miami Beach, it's common for parking lots to charge exhorbitant amounts of money to park, then turn around and have every car still in the lot at 6:30am towed. It's one thing to make somebody pay normal business-hour parking rates for leaving their car parked in the lot past 6:30am. It's another thing entirely to put thousands of moderately drunk individuals in a position where they're forced to choose between a small, but real, risk of getting caught driving with .09 blood alcohol, and the certainty of getting hit with a $250+ towing bill plus everything that goes along with it. Miami Beach allows it, partly because it takes a hefty chunk of all those towing fees and makes a lot more money by practically forcing moderately drunk individuals to risk driving home than it would by encouraging them to leave their cars overnight, take a cab home, and maybe take advantage of being in South Beach the following afternoon to patronize a restaurant.
* Have a police officer present at major parking lots, and offer the following deal if you agree to a breath test before even getting into the car: if you blow over .08, there WILL be a police officer waiting for you around the next corner if you insist upon getting in your car and driving away anyway. If you blow .06 through .08, you'll have two choices: wait until you manage to get it below .06, or be electronically added to the list of drivers flagged for special attention during the next 6 hours (not permanently, but definitely for the next few hours). The grand prize (and reason for consenting) is for drivers who blow less than .06 -- a timestamped certificate that absolves them of any requirement to consent to a breath or blood test for the next 30 minutes if they encounter a checkpoint within 10 miles, 60 minutes if they encounter a checkpoint 10-20 miles away, and 90 minutes if they encounter a checkpoint more than 20 miles away. Essentially, the driver avoids the legal risk of abnormal metabolization, odd readings, or machines with unknown calibration. He's certified as being less than .60, and that's the end of it for the next 30-90 minutes (the distance-dependent time limit is to deal with the case of someone driving to another location and drinking more after being tested... hardcore alcoholics would still slip through that particular crack, but let's be real... someone who's such a hardcore alcoholic that they'd actively try to game the system and sneak additional alcohol after testing lower is going to get caught sooner rather than later, anyway).
> Of course when you combine that with Midsummer snow in Australia
Midsummer *snow*?!? In *Australia?!? From what I've been told, Australia gets snow in the coldest July winter about as often as Orlando or Rome gets snow in January (not entirely unheard of, but people talk about it for years afterward when it happens). Do you really mean it snowed in the middle of Australia's summer? Or do you mean, "It snowed when it was mid-summer in the northern hemisphere, and the coldest dead of winter in Australia and the rest of the southern hemisphere (where seasons are the opposite of those in the northern hemisphere)?"
I guess this falls into the category of "only on Slashdot...", but yeah, it does seem kind of sad that they didn't go for the gold ring at 32,768. It's obviously not a rounding error, either, because then the news story would have said that there are something like 30000, 32000, 33000, or 35000. 32256 -- so close, yet so far... ;-)
127 bottles of of beer on the wall, 127 bottles of beer...
> Most Chinese typists I know use some sort of input means that allows them to type the pinyin (like typing English) and then
> select a character corresponding to the pinyin from a list (done with a number at the end, so the hand never leaves the keyboard).
Correction: most nontechnical and older Chinese users of business applications use Pinyin. The Chinese Slashdot crowd uses a much more efficient input method called Wubizixing that breaks characters down into root radicals and strokes. Wubizixing is more or less the Chinese equivalent of touch-typing QWERTY, and someone who's good at it can type the equivalent of 150 English words per minute.
For key-limited scenarios like cell phones, a variant called Wubihua combines stroke-input (like Wubizixing) with character-picking (like Pinyin), and is actually the input method that's the most usable by westerners who can look at a written character and replicate it, but don't actually know how to pronounce the Mandarin or Cantonese word it represents.
> and for centuries before that Latin was the common language of Europe.
It's important to remember that the Latin taught in classrooms would be mostly unintelligible to most people, including Romans, who actually lived in the Roman Empire at its peak. 2,000 years ago, a Roman Senator might have given speeches in Latin, but the language he actually *spoke* with his family and friends would be mostly recognizable today as the dialect of Italian spoken in southern Italy. A wealthy, powerful Roman Senator from Milan might have written equally lofty Latin, but the language HE spoke among family and friends was strikingly similar to the dialect of Italian spoken today in northern Italy. And so on, as you got farther and farther away from Rome. People in what's now modern Spain, France, and Portugal didn't speak Latin... they spoke languages that were quite similar to modern French, Spanish, and Portuguese. They just pretended that "Latin" was the proper and correct way to represent the language they spoke in writing. Latin didn't magically cease to be the world's common language... people just gradually quit pretending that it was the one and only right way to write their own, and came up with alternate grammar and orthography rules that happened to reflect reality better than Latin did.
> I'd be surprised if the next 20 years didn't see a move towards the international language being Chinese (pinyin).
Actually, I think you're 180 degrees off. If anything, a hundred years from now, English will be even more dominant as a language, but will be commonly (if not officially) written with a few hundred Chinese characters used as alternate ways of writing English words.
First, it'll be done to be "cute" or "cool", then gradually start to be done as an alternate way of abbreviating longer English words with ambiguous abbreviations once an entire generation or two has reached college after growing up with Japanese anime and Chinese sci-fi.
I fully expect to see "" (ren) and "" (nu) on bathroom doors across America by 2030. (with lots of confusion among first-time Chinese visitors, since "" (ren) actually means "people", and will probably lead at least a few visitors to think the men's room is around the corner, and the "" (ren) room is for families with small kids).
Pinyin is passe and 20th-century. Even a semi-clueless westerner who can't speak a word of Mandarin or Cantonese to save his life can type Chinese using wubihua and a numeric keypad after a 10-20 minute lesson on stroke order. In fact, Wubihua has a huge advantage over pinyin -- it's the one input method a westerner can USE if he knows what a character looks like, but doesn't know how to actually say the word it represents.
Now, if only Slashdot would allow Chinese to actually be included in postings without getting mangled...
> Uhm Hindi [wikimedia.org] is the official language of India.
Roughly half of my coworkers were born (and grew up) in India. I think *one* of them might actually be capable of sitting down at a computer running Windows and sending email typed in Hindi. My boss told us point blank that almost nobody in India actually types Hindi, because it's too much of a pain relative to typing English. That's not to say Indian computer users wouldn't *prefer* to have things like menus and text DISPLAYED in Hindi, but as a practical matter, it's almost unheard of for a computer user in India to go to the trouble of switching to Hindi input for something like email, because everyone he or she is likely to care about probably knows English better than the dialect of Hindi the writer would use *anyway*
> Admins upon security advice upload settings which make the device unusable. In that case "reporting compliance"
> while it is not from the user viewpoint is actually a useful feature.
There's actually a useful compromise that's so obvious, it completely blows my mind that it appears to have not even occurred to Microsoft -- keep the corporate data on the server, and give the end users Android and iPhone customized RDP clients that connect to a hosted email app on the server (with the ability to launch intents to alert users to new email when the app isn't actively unlocked, without actually passing sensitive data to the phone itself... ie, "New email" (or maybe "New email from ${name}", if enabled by server policy). Users want to read email? Strong authentication. Phone inactive in "mail" mode for too long? Hide the mail app's view (requiring authentication to re-activate), but leave the rest of the damn phone alone.
This isn't rocket science. Is it bulletproof? Of course not. Any rooted phone can do screenshots. But then again, all you'd need to take an effective screenshot of the most locked-down Android/iPhone ever made is an exotic piece of gear called... (drumroll)... a flatbed scanner. Or a DSLR with macro lens. Or a piece of paper and a pen or pencil. Or a voice recorder. The point is, by keeping the data itself on the server, and using the end user's device only to render the current view, you massively reduce the attack surface... and you get to do it without making the end user completely miserable.