There's a very good reason why tuition skyrocketed over the past ~15 years -- the baby boomers' kids finally hit American colleges and universities like a tsunami, and they showed up expecting the small classes, individual attention, and comfortable surroundings their GenX cousins, aunts, and uncles got to have. See, after the last boomers graduated in the 70s, colleges and universities saw their enrollments PLUMMET. It wasn't quite as visible in states like Florida (that weren't very big during the 60s/70s, and had surging population growth during the 80s and 90s to pick up the slack), but it was VERY visible in states with stable or declining populations, like Ohio, New York, Iowa, Kansas, etc.
To attract students, colleges massively upgraded their amenities, and offered their incoming X'er students a quality of life their Boomer aunts and uncles (remember, most X'ers had parents who were either born DURING WW2 -- a major baby BUST -- or were at the very front end of the postwar baby boom. It cost a bit of money, but for the most part, X'ers were occupying facilities and using infrastructure that was built and paid for MUCH larger student bodies a decade or two earlier.
Fast forward to the late 90s, when the student tsunami washed ashore, and washed ashore HARD. Universities literally couldn't build (and demolish to make room to build) new buildings fast enough. At Florida's public universities, it was an all-out panicked frenzy. 20 years ago, Florida International University in Miami had roughly 5 major buildings, surrounded by an ocean of semi-paved parking lots (the paved parts used to be airport tarmac) on a mostly-empty half-square-mile tract of land. Now, it has buildings and parking garages covering almost all of what used to be the original campus, and has purchased and taken over almost half of the county park that used to be the Dade County fairgrounds (skirting the deeding and zoning issue by using ex-fairgrounds land for things like the new football stadium). The same story is repeated all over Florida... Orlando, Tampa, Jacksonville, Boca Raton, and elsewhere. Schools that used to be small community colleges are now 4-year universities with more students than schools like UCF (University of Central Florida) had in *total* 20 years ago. Santa Fe Community College in Gainesville now has more students than the entire University of Florida did back in the 80s.
Anyway, the point is that all those new buildings cost a veritable shit-ton of money... and most of the universities remember what happened when the flood of boomers ran out -- their enrollment plummeted. Since they couldn't avoid building the new buildings (because it would have meant actively turning away students, and in a state like Florida, would have created an EPIC shortage of admission slots. So, they did the next best thing... they massively jacked up their tuition and fees in an attempt to pay off the bulk of those capital expenditures within ~10 years... before the flood of students and money dries up, and the kids of X'ers start seeping in ~5 years from now to take their place.
Did the easy availability of student loans cause prices to rise? Of course they did. But the very real alternative possibility is that if the universities had NOT undergone that massive expansion, kept their new construction minimal, and limited their incoming students, 80% of GenY kids would have probably been shut out of college completely, and the semi-unique American tradition of mass college education would have become a footnote in history.
Actually, I'd have added two chained pairs of 8-bit shift registers, used them to hold 8-bit Y-b and Y-r chroma data, and encoded it into the msb of the luma data for the previous 16 pixels. In other words, I'd grab 6 bits of data into the upper 6 bits of a register, left-shift the MSB into the LSB of one of the chained pairs of shift registers, then output it as a 6-bit luma with Y-b and Y-r from the OTHER pair of shift registers (defined by the 16 pixels that came before the current 16). The shift registers are chained for in-shifting purposes, but are interpreted independently as two 8-bit Y-b and Y-r registers.
That would give us 640x480 5-bit luma, and 40x480 chroma resolution.
With a bit of extra logic and scratchpad ram, we could "pull a PAL/Secam", and encode the chroma data into adjacent scanlines so we'd end up with 640x480 5-bit luma, and 80x240 chroma resolution.
I've also been intrigued by the possibility of a color encoding scheme based on cyan and red, like the first two-color Technicolor process.Maybe 8 bits per channel (8 for cyan, 8 for red)... or maybe 7 bits per channel, with 2 remaining bits for encoding the "purity" of the cyan channel (00 == cyan as encoded, 01 = a tiny bit more blue than cyan, 10 = a lot more blue than cyan, 11 = pure blue and not cyan at all). In other words, encoding the image like we're "kind of" color blind, but using 2 bits to intelligently nudge it (or shove it) towards pure blue when appropriate.
How, pray tell, can a keylogger flashed "to the BIOS" do ANYTHING meaningful once you've booted into an OS for which the BIOS is (AFAIK) completely irrelevant, like 64-bit Windows 7 or Linux? I mean, I can see how trojan BIOS code might... MIGHT... be able to rewrite the boot sector or something, but the last time I checked, the code in the BIOS is utterly and completely vestigial, irrelevant, and ignored once the Windows logo or 1/2/4 penguins appear and the PC has transitioned to Ring 0 of 386enh mode.
>Diseases are not the only thing that can be cured. Ham, for instance.
Sorry, but the Amiga's most famous graphics mode was congenitally-screwed the moment the company's management forced Jay Miner to go back and make it RGB-based instead of hue-saturation-luminosity based. In retrospect, though, it was mostly just ahead of its time. If I could go back in time 20 years, I'd NOW implement a HAM game by rendering to a phantom 16-bit playfield (using the top bit or few to flag 'dirty' bitmap areas that changed), then use something like the painter algorithm to re-render chunks of it that virtual bitmap to (sliced?)HAM in semi-realtime.
It wouldn't have worked on anything less than an A3000 with at least 2 megs, and would have probably had a real update rate of around.25 to 2fps (not counting sprites), but DAMN, it would have had some killer screenshots in AmigaWorld and sold a few thousand copies before anybody realized the underlying game itself either sucked or was only cool due to the graphics (kind of like the UFO game that was basically a HAM background with sprites animated over it that that sold lots of copies despite sucking as a game, just because it was a game that used HAM).
The problem with relying on backups is the fact that Sandforce SSDs fail so often, and fail in ways that traditional strategies (like RAID and SMART) simply can't deal with. We're talking about drives that would basically commit suicide if your UPS sneezed at the wrong moment, or for other reasons having nothing to do with power loss. Worse, we're talking about drives that can commit "synchronized suicide" and kill themselves in unison if you tried to depend upon RAID 5 or RAID 1 as a strategy for realtime backups.
Sandforce SSDs are SO unreliable (at least, when attempting to emulate a traditional rotational-media hard drive via SATA), they literally aren't safe to use for anything besides a write-through cache that has a "real" hard drive backing it up in realtime. Just be careful if you want to try minimizing the delay by using a SSD as the write-through cache for a 10K Velociraptor... they have lifespan issues of their own.
> Nothing on my hard drive can be lost short of a fairly cataclysmic event that would simultaneous destroy > all copies in existence, and frankly I'd probably be dead then too, so what would I care?
Don't be so sure.
I came TERRIFYINGLY close to losing 20+ years' worth of files permanently last year when my SSD, my Velociraptor-300, AND my 2TB Seagate hard drive all kicked the bucket within a 3-month window of time. At the time, I had the SSD backing itself up to the Velociraptor daily, was backing up the Velociraptor (including the SSD backup) to the 2TB drive weekly, and had the Seagate drive itself backed up to a 3TB external drive once a month or so (the Seagate drive was normally stored at my best friend's house ~15 miles away).
The problem was, as the drives failed and I replaced them, I ended up with multiple copies of recently-modified data, and ended up having a HELL of a time figuring out which was the new and which was the old copy. It took SO LONG to straighten out the resultant mess, that drive #3 ended up failing before I'd finished fully restoring everything from drive #2. And worse, because it took an eternity to do a full backup of the 2TB drive to the external drive (and 4-16 times eternity to restore it), I lost about a month's worth of stuff, and was in cold-sweat panic when I ran out to the store to buy yet another external drive to back up my last surviving copy of the data in case THAT drive failed, too.
Yeah, 2011 was a really, really bad year for my data. In addition to the two external drives, I now also have a complete backup of their contents on ~50 BD-R discs sitting at my parents' house. It took me about a week to burn, and a loss bad enough for me to ever NEED those discs would be devastating... but at least I can sleep at night now knowing that I still have one backup of last resort to fall back on if necessary.
After the crisis, I did a lot of soul-searching and research to find the most robust way to back up my data. What I learned (besides the fact that hard drive reliability has totally gone down the shithole over the past 5 years) was eye-opening.
I'd argue that the SAFEST media for long-term archival backup of files is probably non-LTH BD-R media. It's phase-change magneto-optical, unlike the organic dyes that were the norm for CD-R/RW and DVD+|-/R/RW.
For the record, "LTH" BD-R media uses organic dyes, just like older media, and anecdotal evidence suggests that data written to them has a half-life of approximately 6 months before they start getting correctable errors, and an estimated 18-30 months before they start getting their first uncorrectable errors.
In contrast, most of the phase-change magneto-optical media made by Matsushita and Sony ~15-20 years ago is still readable today (assuming you can find a working drive), and there's no real reason to think BD-R will be any worse (fundamentally, it's the same process now as it was back then... just smaller particles and tighter laser & magnetic fields). In case you're wondering what's magic about them, it's because MO drives use the laser to briefly liquefy the substrate so metallic particles within it can move, and use magnetism to align those particles while it's liquid. Once they re-solidify a moment later, your data is basically "cast in stone" and has no real expiration date.
Incidentally, Millenniata M-disc is basically a DVD-R that's built like a MO BD-R disc. It's one of those cool products that never existed in the format's golden era, but later became possible as a side effect of some newer technology. Kind of like some new gigabit ethernet cards & switches that can also be induced to do 100base-T4 (100mbps ethernet over 4 pairs of cat-3 cable). It was never widely supported back when 100baseT was the norm because it cost too much to add as a feature few cared about, but the technology behind it ended up being used to make gigabit ethernet. Once you have the hardware to do gigabit ethernet, adding retroactive support for 100baseT4 is basically an a
Before everyone gets worked into frothy concern about flash write cycles, keep in mind that the #1 cause of failure & data loss on Sandforce-based controllers is a toxic mess of piss-poor proprietary firmware that's brittle, bitchy, and will brick the drive in self-defense if it corrupts its internal database, then decides you're trying too hard to salvage your data by trying to use something like dd_rescue on it. Oh, and their decision to save a buck by omitting the supercapacitor that's supposed to guarantee that it always has enough power to finish its current write.
Read the Agility/Vertex 2 & 3 forums at ocz if you think I'm making this up. Basically, Sandforce drives have mandatory encryption that can't be disabled to maximize your odds of successful data recovery, but they also employ active countermeasures to detect "hacking attempts" that usually result in the drive ending up in "panic mode".
I wouldn't touch a Sandforce-tainted SSD with a dirty, tetanus-infected pole. They deserve to be sued into oblivion by class-action lawsuits. At the VERY least, they should give us the option of setting our own encryption key (to a value WE know), and a way to rip the bits from a borked drive for offline recovery. The most infuriating thing about data death by Sandforce is the knowledge that 99.99% of your data is *there*, but you aren't allowed to recover it due to their fucked up business policy.
> on top of the short life of their expensive bulbs
WTF are you talking about? I have a Mitsubishi WD-62627 (62", bought in early 2008). It's still on its original bulb, but I finally bought a spare 2 months ago just so I wouldn't have to wait for a replacement when the day finally came. I paid about $40, and got it from Amazon.
Putting it in perspective, my "fragile and expensive bulb" with allegedly-short life has lasted about twice as long as roughly 1/3 of the LCD TVs my friends and family members have purchased over the past 4 years. Most of the LCD TVs sold in America are total and complete garbage, made with parts that die and fail within 2-3 years, and can't be meaningfully repaired because everything in them is proprietary, specific to one or two models, and probably costs more as a replacement part than an entire comparable new TV.
The solution is actually pretty easy and straightforward... get fiber close enough to 99% of Americans to achieve at least 512kbps down and 192kbps up (minimum guaranteed sustained level of service for each end user into the nearest NAP, with the expectation that real-world speeds would be 4-8 times faster, and "end user" defined in a way that guarantees that a house with 4 human inhabitants could get 2mbps down and 768k up of guaranteed aggregate bandwidth into the nearest big-city NAP), and establish a regulatory framework that lowers the barriers to entry for anyone who wants to get connected to that fiber -- ideally, low enough that even an end user willing to throw down a kilobuck or two for hardware could do an end-run around everyone standing in the way and put his own fixed wireless antenna & transceiver on the (possibly) government-owned tower at the nearest fiber point if necessary. Then resell access to his adjacent neighbors if he felt like it.
Allow companies like AT&T and Comcast to buy wholesale access to that fiber and lay their own fiber, coax, or even just repurpose copper pairs... but ensure that if push came to shove, end users could show them their middle finger, leapfrog over them at their own expense, and peer into the government-laid fiber on the same neutral terms as AT&T or Comcast themselves. Where possible, the government could encourage those big companies to lay the fiber instead, and give them low-interest long-term loans that are contingent upon them offering access to end users on the same neutral common-carrier terms... but also giving the FCC (or its delegate) the authority to seize and operate that infrastructure directly if the company becomes unwilling or unable to (kind of like how the FDIC can seize a bank when it becomes insolvent, and operate it directly under receivership to prevent disruptions for customers who depend on it.
The government shouldn't try to be the ISP of last resort for poor Americans, but it SHOULD strive to eliminate what would otherwise be insurmountable structural barriers to end users who fall through the cracks of AT&T/Comcast's business plans, and lower the bar enough to allow highly-motivated end users to take matters into their own hands without having to depend upon AT&T or Comcast for the solution.
I think it's more a case that the free market is pretty crap at SUSTAINED, ONGOING research. I mean, we've all seen the pattern so many times, it's practically a meme... someone comes up with something groundbreaking and innovative, starts a company, gets major investors, becomes a major force... then the company begins its long trajectory downward... never really coming up with anything equally innovative -- or even halfway impressive -- again.
TiVo? Groundbreaking and innovative 10 years ago... but now? Meh. Basically, the same as 10 years ago, except the hard drives are bigger and they can do HD. Wheeeeeee.
Windows? Win95 rocked compared to pretty much everything that came before it. For the first time in history, my Ultrasound, my ET4000w32 card, and my funky caching VL-bus hard drive controller with a few megs of ram onboard all basically worked. Fast forward to the 21st century... Windows 8. If Microsoft announced they were giving it away free via Windows Update next week, most of us would unplug our DSL & cable modems, and probably physically remove the hard drives from our computers and store them in foil bags in a vault for 2 weeks... *just* to be safe. After backing them up, disabling Windows Update, and explicitly firewalling anything at *.microsoft.com and Microsoft's/8 block of IP addresses forever going forward.
Sirius? I loved it. It saved my sanity after my two favorite radio stations in Miami both got destroyed over the span of 3 days. Then... well... the merger happened. Their audio quality went down the shithole, their channels became more like XM, and less like popular big-city radio stations that just didn't have commercials.
Computer hardware. For the past 5 years or so, it's basically stagnated... if not regressed. The i7 has gained a whopping 400MHz or so. And chiclet keyboards. Godfuckingdamn chiclet keyboards. Even on Thinkpads. THINKPADS! Jesus God naked on a motorcycle with a buttplug. THINKPADS! (goes off to cry...)
Just to name a few companies... let's start with General Electric. How many years has it been since they've innovated anything more meaningful than finding a cheaper factory in China to make the same clock radios they've been making for the past 20 years? *DO* they even invent anything anymore?
Or Disney. Remember when Disney was the company that built monorails and everything they did was magic? Now they run buses to their parks from cookie-cutter hotels that might as well be a Holiday Inn Express. They've made god knows how many sequels to Toy Story & Cars, and shamelessly prostituted movies from our childhood that were once sacred and special. And now they own Star Wars. May the Force help us all.
Oh, I forgot the crowning glory. Mrs. Butterworth syrup. Two percent butter. Yum! Then 1-1/2%. Then 1%. Then 1/2%. Then "flavored with". Then "Buttery". Bastards.Another piece of America shamelessly destroyed.
Yes there was. Microsoft finally gave us real symlinks, so we no longer had to kludge them with NTFS junctions.
> If Metro is such a problem. Buy Windows 8, Install Start8 and never worry about metro again
And how, pray tell, do you restore Aero glass, proper handling of multiple monitors, and all the other little refinements that Microsoft *finally* got working again in Win7 after Vista broke them, then the preview releases of Win8 apparently took away again?
The fundamental problem with Windows 8 is that Microsoft took everything it's learned over the past 20 years about usability with high-end mouse-driven computers with multiple hi-res displays, and flushed it all down the toilet so phone apps can pretend they're real Windows applications worthy of respect instead of third-rate substitutes for the real thing. In effect, Microsoft's "solution" to the problem of desktop apps sucking heinously when run on phones with tiny touchscreens was to abolish desktop apps.
I desperately want you to be right, but the truth is, eBook readers still largely suck in terms of random-access flip-through usability compared to... well.. a real book. The sad truth is, it's just not the same. I've bought about a dozen ebooks so far, and none of them have really been read much. They feel more like curated blog postings and online tutorials than like real books.
Plus, it's all the little things wrong. Formatting glitches nobody bothered to fix. Bad page breaks. Pages whose composition nobody really gave a shit about, where a real book would have had somebody at the publishing house agonizing over and tweaking the layout for a day. eBooks feel raw and perpetually unfinished, even when they *aren't* "early access" copies. It's like, everyone knows they're going to do further editing, so nobody at the publisher wants to invest the time making it look good *today*, then when "tomorrow" comes, nobody has time to make it look polished *then*, either, because the book is "done", most of the people who are ever going to buy it already paid for it pre-release, and there are 10 more books they have to work on to bring in more revenue instead. Real books have a permanence about them that motivates publishers to spend the extra time polishing their layout, tweaking the kerning, and all the other little things that Make a Difference to the overall reading experience.
Did I mention the fact that eReaders pretty much universally suck for non-sequential reading workflows? Kindle might be a net improvement over a 1,000 page printed novel, but it's not an improvement over a random printed Manning book. Every e-reader I've seen feels like you're wading through wet concrete when flipping through the pages.
And dammit, I'm still trying to find an Android pdf or epub reader that will let me read 2-up pages (like an open book) on my Xoom in landscape orientation (intelligently trimming away much of the margins, but maintaining the relative layout and flow of the printed book).
The other problem is that the hardware is just too underpowered. Books might be static while you're reading, but replicating the experience of flipping pages in a real book requires some major CPU power... not to mention fast, random-access media and lots of ram. None of this "fetch a tiny chunk at a time through a microSD single-bit cocktail straw" and "run at 200MHz to make the undersized battery last more than 2 hours" bullshit.
Verizon and Sprint had no choice but to upgrade (to LTE or Wimax), because EVDO rev.A hits a brick wall at ~2mbps. They COULD have upgraded to EVDO.revB, but the capital cost would have been almost the same, and they would have ended up locked in to Qualcomm as a single-source vendor forever, and paid premium prices for everything they bought going forward.
For AT&T and T-Mobile, the benefits of LTE aren't nearly as big. They're real, and they exist, but they aren't earth-shaking for 80% or so of their real-world users -- the 80% for whom HSPA+ already gives 16-20mbps, or for whom 1700MHz LTE would be unusable anyway. Remember, real-world HSPA is ~4-6mbps, and most HSPA+ is 10-16mbps. LTE might start at 16mbps and creep up to 26-30mbps (limited mainly by backhaul fiber capacity), but LTE's brick wall is signal strength -- below a certain point that's ~10dBm higher than the minimum needed by HSPA(+), LTE doesn't work *at all*.
I'm not 100% sure, but I believe that in a moving vehicle on a freeway (or a train), HSPA+ through urban areas might even work better and more consistently than LTE, simply because the phone can (theoretically) maintain at least one active tower connection at all times while leapfrogging between alternating streams. So... you might have tower 1 and 2 actively connected, lose tower 2, connect to tower 3, lose tower 1, connect to tower 4, lose tower 3, connect to tower 5, and so on... falling to 4-6mbps, bursting to 10-12mbps, and repeating over and over again, but always remaining connected by at least one or the other. In contrast, with LTE, you'd have periods of time with no connection at all when it broke the current connection to establish a new one with the next tower.
T-Mobile is deploying LTE, but in all honesty, the distinction between them is moot for probably 70-80% of their customers.
HSPA+, when it works properly, is basically as fast as LTE. The catch is, HSPA+ only gets its fastest rates if the user has solid connections to two or more towers, because it works the same way a dialup shotgun modem worked -- the phone independently connects to two or more towers, then splits the bitstream and sends part to each tower. Upstream, the bits are recombined into a single bitstream by T-Mobile's network.
Where LTE makes a difference is suburbia. Specifically, suburban locations where the user can only see one tower well, but has a rock-solid signal from that one tower. THEN, the user might get 16-20mbps from LTE, but only 4-6mbps from HSPA.
Here's the catch: if the user has a mediocre signal from two towers, but a strong signal from NONE, he might discover that T-Mobile's 1700MHz LTE doesn't work at all, and he's still limited to 10-12mbps HSPA+. If the user has a mediocre signal from ONE tower, he'll probably be lucky to see 1-2mbps, just like he does now, and LTE won't work either.
Why? LTE's throughput is kind of like 8-VSB TV transmissions. When it's strong enough, it's flawless and full-speed, even when other radio modes are degrading badly. But the moment your signal gets even a tiny bit weaker than the minimum (roughly 10dBm stronger than the minimum for viable HSPA), you fall off the cliff and lose it entirely. You don't get slower LTE... you have no LTE *whatsoever*.
So... one strong tower == LTE faster than HSPA(+)
Two mediocre towers == HSPA+ mostly works, LTE doesn't work reliably.
One mediocre tower == HSPA limps along, LTE doesn't work at all.
Two or more strong towers == LTE slightly faster than HSPA+ in theory, and might use less power, but looking at speed alone, you'd be hard-pressed to tell with any real certainty whether the user was using LTE or HSPA+. LTE has better latency (no need to demux, split, buffer, and recombine bitstreams, and the symbol rate itself is faster), but the difference isn't all that huge.
A 10kW generator is *barely* enough to run 2 or 3 10,000BTU window air conditioners. I believe you need ~24kW of 220v capacity to start a normal 2-3 ton central air conditioner.
It's a shame companies like Carrier, Rheem, etc can't put a little effort into designing central ac units that are "generator friendly" & can start with less inrush current. Like, maybe some kind of transmission that would allow the compressor to spin up slowly, instead of just soaking up 20+ kW for 3 seconds before settling down to half that amount. Or logic to start up the compressor, THEN the blower fan, instead of both at once.
Ummm... try testing that theory by tossing a chunk of elemental sodium or calcium into a bucket of water (or swimming pool, or jacuzzi). There are plenty of videos on Youtube illustrating the outcome...
AT&T (in my case, U-verse) might have an annoying cap, but it's big enough to not really bother me much. On the other had, my AT&T wireless cap (3 gigabytes) is terrifyingly low, because unlike the former cap, there's a real chance I could go over it and get charged extra in any given month. So I *do* actually care about the details of how they calculate it.
Specifically, is AT&T Wireless billing "chunky"? Suppose I have an Andriod app running in the background that relentlessly polls some remote server every 10 seconds. Or 60 seconds, if it matters. My hypothetical app tries to keep the bandwidth down, and works as follows:
User sends 1-byte command via UDP, then disconnects.
Server looks at the request to discern the sender's IP and port, decides what to send based on the command byte, and sends a one-byte response.
As far as I know, a UDP datagram with a single-byte payload is 33 bytes long (24 bytes for the IP header, 8 bytes for the UDP header, 1 byte for the payload). It's just a hunch, but I suspect that my likelihood of getting billed for exactly 66 bytes of data use is 100% pure fantasy. I'd be shocked if those two datagrams didn't end up getting billed as two 1k chunks of data, or worse.
Are wireless carriers required to file explicit tariffs with state utility regulators disclosing their exact charges, and explain how they calculate any usage-based fees in detail?
Yes. If the US has a shitload of shale, China and Russia almost certainly do, too & just haven't found it yet. And I think it's a bit premature to think their response will be to say, "This isn't Clean Green Energy(tm), so we mustn't be like those naughty Earth-hating Americans and use it."
Far MORE likely is that average European voters will get pissed off about higher energy prices in Europe, eventually elect leaders who aren't quite as determined to be Green at any cost, and join the celebration. And probably discover plenty of their own shale while they're at it.
Not even the most wildly-optimistic supporter of "green energy" fantasizes that it can produce as much cheap energy as fossil fuels and nuclear power. Their plans all involve "somehow" imposing life with less energy on the rest of the world.
In the long run, if you want to encourage the use of fuels that don't produce greenhouse gases, support nuclear power. It's far from perfect, and can be nasty stuff at times, but the truth is... it's the best thing we have, and the best thing we're likely to have for a really long time. If we reprocessed intensely-radioactive spent fuel instead of trying to store mass quantities of it on-site (with long-term plans to bury it under a mountain somewhere), the waste problem would mostly take care of itself.
If you think of spent nuclear fuel like Alkaline batteries, reprocessing is kind of like adding a "joule thief" circuit (which multiplies the voltage of a dying battery, even as it falls off the voltage cliff) and DC-DC regulator to a consumer electronics device that would normally just use 4 AA batteries to grab ~4.5 to 6 volts, so you can wring useful power out of the battery until it's literally squeezed dry and barely outputting enough DC to detect with a multimeter.
> (...snip...) We find some dirty old bag of crack (...snip...)
"We clear out the trash that's been down in the basement since we moved in, and discover an old, forgotten sub-basement built in the 19th century for coal that's filled with enough dirty old crack to to fill our own habit, plus the habits of the entire northeastern US, for the next 5000 years, and scream "Yeah! We're saved!..."
There are two things you never, ever, EVER want on your main webserver or app server: blog software (Wordpress, etc) and user-generated or uploaded content. Both are several orders of magnitude more likely to get your server pwn3d, and keeping them at arm's length from your main site & app server is kind of like building a ship with bulkhead doors to contain flooding in the event of a breach.
I speak from experience. Wordpress got totally owned somehow on its server. My app? Safe, 2500 miles away, on a server with no usernames/passwords/hostnames in common.
The problem is that apps like VBulletin, PhpBB, & others might actually be more secure than YOUR app, but your app doesn't have 47,000 hackers searching it for vulnerabilities & ready to exploit every site using the same software the moment a 0-day is found. Web forums are "high-value" targets. Your app, running one one server with a few thousand daily users, is not. You'll still get attacked, of course, but you're less likely to get compromised by some wacky, obscure bug in a system library that only affects a few servers (including yours), because THOSE exploits are only worth the time & effort of attackers when they can be automated or have enough potential victims to be worthwhile.
Another reason for a totally different hostname: to keep your whole site from getting blacklisted by Google over an XSS exploit some user pulled off in a posting.
The problem is that in the US, you (or a corporation) can't take a patent owner to court (seeking to overturn his patent) unless you/the corporation have "standing" -- ie, an infringement suit has been filed against you. And the patent's owner can drop the suit against you at any time up to the final moment the judge rules it invalid (preventing the troll from suing YOU again, but not others). I might be wrong, but I believe that in China, it *is* possible for a coalition of likely victims to gang up on the patent's owner & sue to get it overturned.
In the US, you CAN seek "re-examination", but it's expensive, the odds are overwhelmingly stacked against you, and (afaik), the only thing "winning" gets for you is immunity from treble damages if you act upon your victory, get sued for infringement anyway, and they prevail.
The huge bug in the US patent system is that patents are presumed valid unless overturned in court, but courts operate on the assumption that patents wouldn't have been granted without merit, and there have been plenty of instances (particularly the late 90s/early 00s) when overwhelmed examiners approved questionable patents on the theory that the courts could sort out their mess later.
The problem is compounded by American patent law's willingness to grant total blocking power to even the most incidental case of infringement. IE, if I hold the first patent on some primitive implementation that barely even works (if it works at all), and you *profoundly* improve upon it and make it viable, I can still block you from using it because my patent covers something more fundamental. I'm pretty sure that in China, there's a statutory process for compulsory licensing where someone who has the profoundly-improved patent could file the paperwork, then tell the owner of patent #1, "Shut up, take your fair share of the money, and file a lawsuit if you think you deserve more."
In the US, patent owners have too much power to prevent their patent's use by others at *any* cost, sane or otherwise.
There's no legal process to say, "Yes, I agree his patent is valid, but he's an asshole/troll/sociopath and is demanding more than its fair market value relative to the whole product, or is flat-out refusing to licence it at all." So, we have cases of companies restricting patent use to force you to go with one service provider or another -- RESTRICTING innovation, forcing consumers into false dichotomies, and preventing them from being able to have ideal devices that combine the best of everything -- from Edison's hated, inferior stock tickers to TiVo-vs-Dish/DirecTV/AT&T (cablecards don't work with them, so you CAN'T buy and use your own DVR) and Apple-vs-Samsung.
Moore's law says nothing about speed. Moore's law is about the number of transistors.
Netburst (a.k.a., the Pentium 4) was a miserable, failed experiment in CPU Marketecture (optimizing the wrong features just to give the marketing department big, meaningless numbers to dangle in front of consumers who don't know any better). It was designed from the start to offer cheap, worthless gigahertz, with less real performance than older CPUs running at half their nominal speed. May it rot in hell (and help keep it toasty and warm) forever.
In historical terms, the Pentium 4 family was kind of like trying to make freight trains faster by bolting a jet engine on top of a boxcar and using its exhaust to propel it. The Core family is metaporically like someone realizing they'd be better off using the jet engine to drive a turbine generator in the locomotive, buffer the power with a large bank of batteries, and use it to run bigger electric motors to push and pull the train from both ends instead.
Tell me about it. I have a nominally-1.5GHz quadcore Android phone that, when running Graffiti, can barely tell the difference between a "G" drawn like a "6" on the letter side, and the letter "O" with better than 90% accuracy unless I use SetCPU to lock it to full speed (with devastating impact upon battery life) whenever the screen is on, yet somehow... SOMEHOW... a slow, lowly 16MHz Dragonball m68k could do the same thing with nearly perfect, flawless accuracy. The biggest single reason, as far as I can tell? On the Palm, the digitizer was read via timer interrupt with clockwork regularity, and the recognition routine was 100% assembly. I think the Android version was written in Java, gets to have a sample whenever Android feels like allowing it to have one, and has its accuracy flushed down the toilet by endless speed-scaling that destroys the motion-vector algorithm that made the original so good, then made even worse by the insistence of Google (and Amazon, and everyone else) upon making AJAX network calls between every goddamn letter, with no real way to disable it (there used to be... briefly... but Google appears to have taken it away for good at some point between Gingerbread and ICS).
Sigh. Someday, I'm going to finally learn how to implement kernel-level Graffiti on Android (well, ok... kernel-level digitizer sampling with clockwork, non-scaled regularity that gets neatly buffered, so that the higher-level recognition routine can lurch and stumble as usual, but at least it'll have solid sample data to generate its motion vectors from).
Slashdot Mirror
There's a very good reason why tuition skyrocketed over the past ~15 years -- the baby boomers' kids finally hit American colleges and universities like a tsunami, and they showed up expecting the small classes, individual attention, and comfortable surroundings their GenX cousins, aunts, and uncles got to have. See, after the last boomers graduated in the 70s, colleges and universities saw their enrollments PLUMMET. It wasn't quite as visible in states like Florida (that weren't very big during the 60s/70s, and had surging population growth during the 80s and 90s to pick up the slack), but it was VERY visible in states with stable or declining populations, like Ohio, New York, Iowa, Kansas, etc.
To attract students, colleges massively upgraded their amenities, and offered their incoming X'er students a quality of life their Boomer aunts and uncles (remember, most X'ers had parents who were either born DURING WW2 -- a major baby BUST -- or were at the very front end of the postwar baby boom. It cost a bit of money, but for the most part, X'ers were occupying facilities and using infrastructure that was built and paid for MUCH larger student bodies a decade or two earlier.
Fast forward to the late 90s, when the student tsunami washed ashore, and washed ashore HARD. Universities literally couldn't build (and demolish to make room to build) new buildings fast enough. At Florida's public universities, it was an all-out panicked frenzy. 20 years ago, Florida International University in Miami had roughly 5 major buildings, surrounded by an ocean of semi-paved parking lots (the paved parts used to be airport tarmac) on a mostly-empty half-square-mile tract of land. Now, it has buildings and parking garages covering almost all of what used to be the original campus, and has purchased and taken over almost half of the county park that used to be the Dade County fairgrounds (skirting the deeding and zoning issue by using ex-fairgrounds land for things like the new football stadium). The same story is repeated all over Florida... Orlando, Tampa, Jacksonville, Boca Raton, and elsewhere. Schools that used to be small community colleges are now 4-year universities with more students than schools like UCF (University of Central Florida) had in *total* 20 years ago. Santa Fe Community College in Gainesville now has more students than the entire University of Florida did back in the 80s.
Anyway, the point is that all those new buildings cost a veritable shit-ton of money... and most of the universities remember what happened when the flood of boomers ran out -- their enrollment plummeted. Since they couldn't avoid building the new buildings (because it would have meant actively turning away students, and in a state like Florida, would have created an EPIC shortage of admission slots. So, they did the next best thing... they massively jacked up their tuition and fees in an attempt to pay off the bulk of those capital expenditures within ~10 years... before the flood of students and money dries up, and the kids of X'ers start seeping in ~5 years from now to take their place.
Did the easy availability of student loans cause prices to rise? Of course they did. But the very real alternative possibility is that if the universities had NOT undergone that massive expansion, kept their new construction minimal, and limited their incoming students, 80% of GenY kids would have probably been shut out of college completely, and the semi-unique American tradition of mass college education would have become a footnote in history.
Actually, I'd have added two chained pairs of 8-bit shift registers, used them to hold 8-bit Y-b and Y-r chroma data, and encoded it into the msb of the luma data for the previous 16 pixels. In other words, I'd grab 6 bits of data into the upper 6 bits of a register, left-shift the MSB into the LSB of one of the chained pairs of shift registers, then output it as a 6-bit luma with Y-b and Y-r from the OTHER pair of shift registers (defined by the 16 pixels that came before the current 16). The shift registers are chained for in-shifting purposes, but are interpreted independently as two 8-bit Y-b and Y-r registers.
That would give us 640x480 5-bit luma, and 40x480 chroma resolution.
With a bit of extra logic and scratchpad ram, we could "pull a PAL/Secam", and encode the chroma data into adjacent scanlines so we'd end up with 640x480 5-bit luma, and 80x240 chroma resolution.
I've also been intrigued by the possibility of a color encoding scheme based on cyan and red, like the first two-color Technicolor process.Maybe 8 bits per channel (8 for cyan, 8 for red)... or maybe 7 bits per channel, with 2 remaining bits for encoding the "purity" of the cyan channel (00 == cyan as encoded, 01 = a tiny bit more blue than cyan, 10 = a lot more blue than cyan, 11 = pure blue and not cyan at all). In other words, encoding the image like we're "kind of" color blind, but using 2 bits to intelligently nudge it (or shove it) towards pure blue when appropriate.
How, pray tell, can a keylogger flashed "to the BIOS" do ANYTHING meaningful once you've booted into an OS for which the BIOS is (AFAIK) completely irrelevant, like 64-bit Windows 7 or Linux? I mean, I can see how trojan BIOS code might... MIGHT... be able to rewrite the boot sector or something, but the last time I checked, the code in the BIOS is utterly and completely vestigial, irrelevant, and ignored once the Windows logo or 1/2/4 penguins appear and the PC has transitioned to Ring 0 of 386enh mode.
>Diseases are not the only thing that can be cured. Ham, for instance.
Sorry, but the Amiga's most famous graphics mode was congenitally-screwed the moment the company's management forced Jay Miner to go back and make it RGB-based instead of hue-saturation-luminosity based. In retrospect, though, it was mostly just ahead of its time. If I could go back in time 20 years, I'd NOW implement a HAM game by rendering to a phantom 16-bit playfield (using the top bit or few to flag 'dirty' bitmap areas that changed), then use something like the painter algorithm to re-render chunks of it that virtual bitmap to (sliced?)HAM in semi-realtime.
It wouldn't have worked on anything less than an A3000 with at least 2 megs, and would have probably had a real update rate of around .25 to 2fps (not counting sprites), but DAMN, it would have had some killer screenshots in AmigaWorld and sold a few thousand copies before anybody realized the underlying game itself either sucked or was only cool due to the graphics (kind of like the UFO game that was basically a HAM background with sprites animated over it that that sold lots of copies despite sucking as a game, just because it was a game that used HAM).
The problem with relying on backups is the fact that Sandforce SSDs fail so often, and fail in ways that traditional strategies (like RAID and SMART) simply can't deal with. We're talking about drives that would basically commit suicide if your UPS sneezed at the wrong moment, or for other reasons having nothing to do with power loss. Worse, we're talking about drives that can commit "synchronized suicide" and kill themselves in unison if you tried to depend upon RAID 5 or RAID 1 as a strategy for realtime backups.
Sandforce SSDs are SO unreliable (at least, when attempting to emulate a traditional rotational-media hard drive via SATA), they literally aren't safe to use for anything besides a write-through cache that has a "real" hard drive backing it up in realtime. Just be careful if you want to try minimizing the delay by using a SSD as the write-through cache for a 10K Velociraptor... they have lifespan issues of their own.
> Nothing on my hard drive can be lost short of a fairly cataclysmic event that would simultaneous destroy
> all copies in existence, and frankly I'd probably be dead then too, so what would I care?
Don't be so sure.
I came TERRIFYINGLY close to losing 20+ years' worth of files permanently last year when my SSD, my Velociraptor-300, AND my 2TB Seagate hard drive all kicked the bucket within a 3-month window of time. At the time, I had the SSD backing itself up to the Velociraptor daily, was backing up the Velociraptor (including the SSD backup) to the 2TB drive weekly, and had the Seagate drive itself backed up to a 3TB external drive once a month or so (the Seagate drive was normally stored at my best friend's house ~15 miles away).
The problem was, as the drives failed and I replaced them, I ended up with multiple copies of recently-modified data, and ended up having a HELL of a time figuring out which was the new and which was the old copy. It took SO LONG to straighten out the resultant mess, that drive #3 ended up failing before I'd finished fully restoring everything from drive #2. And worse, because it took an eternity to do a full backup of the 2TB drive to the external drive (and 4-16 times eternity to restore it), I lost about a month's worth of stuff, and was in cold-sweat panic when I ran out to the store to buy yet another external drive to back up my last surviving copy of the data in case THAT drive failed, too.
Yeah, 2011 was a really, really bad year for my data. In addition to the two external drives, I now also have a complete backup of their contents on ~50 BD-R discs sitting at my parents' house. It took me about a week to burn, and a loss bad enough for me to ever NEED those discs would be devastating... but at least I can sleep at night now knowing that I still have one backup of last resort to fall back on if necessary.
After the crisis, I did a lot of soul-searching and research to find the most robust way to back up my data. What I learned (besides the fact that hard drive reliability has totally gone down the shithole over the past 5 years) was eye-opening.
I'd argue that the SAFEST media for long-term archival backup of files is probably non-LTH BD-R media. It's phase-change magneto-optical, unlike the organic dyes that were the norm for CD-R/RW and DVD+|-/R/RW.
For the record, "LTH" BD-R media uses organic dyes, just like older media, and anecdotal evidence suggests that data written to them has a half-life of approximately 6 months before they start getting correctable errors, and an estimated 18-30 months before they start getting their first uncorrectable errors.
In contrast, most of the phase-change magneto-optical media made by Matsushita and Sony ~15-20 years ago is still readable today (assuming you can find a working drive), and there's no real reason to think BD-R will be any worse (fundamentally, it's the same process now as it was back then... just smaller particles and tighter laser & magnetic fields). In case you're wondering what's magic about them, it's because MO drives use the laser to briefly liquefy the substrate so metallic particles within it can move, and use magnetism to align those particles while it's liquid. Once they re-solidify a moment later, your data is basically "cast in stone" and has no real expiration date.
Incidentally, Millenniata M-disc is basically a DVD-R that's built like a MO BD-R disc. It's one of those cool products that never existed in the format's golden era, but later became possible as a side effect of some newer technology. Kind of like some new gigabit ethernet cards & switches that can also be induced to do 100base-T4 (100mbps ethernet over 4 pairs of cat-3 cable). It was never widely supported back when 100baseT was the norm because it cost too much to add as a feature few cared about, but the technology behind it ended up being used to make gigabit ethernet. Once you have the hardware to do gigabit ethernet, adding retroactive support for 100baseT4 is basically an a
Before everyone gets worked into frothy concern about flash write cycles, keep in mind that the #1 cause of failure & data loss on Sandforce-based controllers is a toxic mess of piss-poor proprietary firmware that's brittle, bitchy, and will brick the drive in self-defense if it corrupts its internal database, then decides you're trying too hard to salvage your data by trying to use something like dd_rescue on it. Oh, and their decision to save a buck by omitting the supercapacitor that's supposed to guarantee that it always has enough power to finish its current write.
Read the Agility/Vertex 2 & 3 forums at ocz if you think I'm making this up. Basically, Sandforce drives have mandatory encryption that can't be disabled to maximize your odds of successful data recovery, but they also employ active countermeasures to detect "hacking attempts" that usually result in the drive ending up in "panic mode".
I wouldn't touch a Sandforce-tainted SSD with a dirty, tetanus-infected pole. They deserve to be sued into oblivion by class-action lawsuits. At the VERY least, they should give us the option of setting our own encryption key (to a value WE know), and a way to rip the bits from a borked drive for offline recovery. The most infuriating thing about data death by Sandforce is the knowledge that 99.99% of your data is *there*, but you aren't allowed to recover it due to their fucked up business policy.
> on top of the short life of their expensive bulbs
WTF are you talking about? I have a Mitsubishi WD-62627 (62", bought in early 2008). It's still on its original bulb, but I finally bought a spare 2 months ago just so I wouldn't have to wait for a replacement when the day finally came. I paid about $40, and got it from Amazon.
Putting it in perspective, my "fragile and expensive bulb" with allegedly-short life has lasted about twice as long as roughly 1/3 of the LCD TVs my friends and family members have purchased over the past 4 years. Most of the LCD TVs sold in America are total and complete garbage, made with parts that die and fail within 2-3 years, and can't be meaningfully repaired because everything in them is proprietary, specific to one or two models, and probably costs more as a replacement part than an entire comparable new TV.
The solution is actually pretty easy and straightforward... get fiber close enough to 99% of Americans to achieve at least 512kbps down and 192kbps up (minimum guaranteed sustained level of service for each end user into the nearest NAP, with the expectation that real-world speeds would be 4-8 times faster, and "end user" defined in a way that guarantees that a house with 4 human inhabitants could get 2mbps down and 768k up of guaranteed aggregate bandwidth into the nearest big-city NAP), and establish a regulatory framework that lowers the barriers to entry for anyone who wants to get connected to that fiber -- ideally, low enough that even an end user willing to throw down a kilobuck or two for hardware could do an end-run around everyone standing in the way and put his own fixed wireless antenna & transceiver on the (possibly) government-owned tower at the nearest fiber point if necessary. Then resell access to his adjacent neighbors if he felt like it.
Allow companies like AT&T and Comcast to buy wholesale access to that fiber and lay their own fiber, coax, or even just repurpose copper pairs... but ensure that if push came to shove, end users could show them their middle finger, leapfrog over them at their own expense, and peer into the government-laid fiber on the same neutral terms as AT&T or Comcast themselves. Where possible, the government could encourage those big companies to lay the fiber instead, and give them low-interest long-term loans that are contingent upon them offering access to end users on the same neutral common-carrier terms... but also giving the FCC (or its delegate) the authority to seize and operate that infrastructure directly if the company becomes unwilling or unable to (kind of like how the FDIC can seize a bank when it becomes insolvent, and operate it directly under receivership to prevent disruptions for customers who depend on it.
The government shouldn't try to be the ISP of last resort for poor Americans, but it SHOULD strive to eliminate what would otherwise be insurmountable structural barriers to end users who fall through the cracks of AT&T/Comcast's business plans, and lower the bar enough to allow highly-motivated end users to take matters into their own hands without having to depend upon AT&T or Comcast for the solution.
I, for one, am glad our government has set such lofty goals, because I really want a RTG power pack for my cell phone. ( http://en.wikipedia.org/wiki/Radioisotope_thermoelectric_generator )
I think it's more a case that the free market is pretty crap at SUSTAINED, ONGOING research. I mean, we've all seen the pattern so many times, it's practically a meme... someone comes up with something groundbreaking and innovative, starts a company, gets major investors, becomes a major force... then the company begins its long trajectory downward... never really coming up with anything equally innovative -- or even halfway impressive -- again.
TiVo? Groundbreaking and innovative 10 years ago... but now? Meh. Basically, the same as 10 years ago, except the hard drives are bigger and they can do HD. Wheeeeeee.
Windows? Win95 rocked compared to pretty much everything that came before it. For the first time in history, my Ultrasound, my ET4000w32 card, and my funky caching VL-bus hard drive controller with a few megs of ram onboard all basically worked. Fast forward to the 21st century... Windows 8. If Microsoft announced they were giving it away free via Windows Update next week, most of us would unplug our DSL & cable modems, and probably physically remove the hard drives from our computers and store them in foil bags in a vault for 2 weeks... *just* to be safe. After backing them up, disabling Windows Update, and explicitly firewalling anything at *.microsoft.com and Microsoft's /8 block of IP addresses forever going forward.
Sirius? I loved it. It saved my sanity after my two favorite radio stations in Miami both got destroyed over the span of 3 days. Then... well... the merger happened. Their audio quality went down the shithole, their channels became more like XM, and less like popular big-city radio stations that just didn't have commercials.
Computer hardware. For the past 5 years or so, it's basically stagnated... if not regressed. The i7 has gained a whopping 400MHz or so. And chiclet keyboards. Godfuckingdamn chiclet keyboards. Even on Thinkpads. THINKPADS! Jesus God naked on a motorcycle with a buttplug. THINKPADS! (goes off to cry...)
Just to name a few companies... let's start with General Electric. How many years has it been since they've innovated anything more meaningful than finding a cheaper factory in China to make the same clock radios they've been making for the past 20 years? *DO* they even invent anything anymore?
Or Disney. Remember when Disney was the company that built monorails and everything they did was magic? Now they run buses to their parks from cookie-cutter hotels that might as well be a Holiday Inn Express. They've made god knows how many sequels to Toy Story & Cars, and shamelessly prostituted movies from our childhood that were once sacred and special. And now they own Star Wars. May the Force help us all.
Oh, I forgot the crowning glory. Mrs. Butterworth syrup. Two percent butter. Yum! Then 1-1/2%. Then 1%. Then 1/2%. Then "flavored with". Then "Buttery". Bastards.Another piece of America shamelessly destroyed.
> There was nothing redeeming about that OS.
Yes there was. Microsoft finally gave us real symlinks, so we no longer had to kludge them with NTFS junctions.
> If Metro is such a problem. Buy Windows 8, Install Start8 and never worry about metro again
And how, pray tell, do you restore Aero glass, proper handling of multiple monitors, and all the other little refinements that Microsoft *finally* got working again in Win7 after Vista broke them, then the preview releases of Win8 apparently took away again?
The fundamental problem with Windows 8 is that Microsoft took everything it's learned over the past 20 years about usability with high-end mouse-driven computers with multiple hi-res displays, and flushed it all down the toilet so phone apps can pretend they're real Windows applications worthy of respect instead of third-rate substitutes for the real thing. In effect, Microsoft's "solution" to the problem of desktop apps sucking heinously when run on phones with tiny touchscreens was to abolish desktop apps.
> The only thing you could buy in the 90ies that'd last you a lifetime is a Soundblaster 16.
I guess you've never heard of the IBM Model M keyboard. I fully expect to STILL be using mine 25 years from now. Or at least, one of my 8 spares ;-)
I desperately want you to be right, but the truth is, eBook readers still largely suck in terms of random-access flip-through usability compared to... well.. a real book. The sad truth is, it's just not the same. I've bought about a dozen ebooks so far, and none of them have really been read much. They feel more like curated blog postings and online tutorials than like real books.
Plus, it's all the little things wrong. Formatting glitches nobody bothered to fix. Bad page breaks. Pages whose composition nobody really gave a shit about, where a real book would have had somebody at the publishing house agonizing over and tweaking the layout for a day. eBooks feel raw and perpetually unfinished, even when they *aren't* "early access" copies. It's like, everyone knows they're going to do further editing, so nobody at the publisher wants to invest the time making it look good *today*, then when "tomorrow" comes, nobody has time to make it look polished *then*, either, because the book is "done", most of the people who are ever going to buy it already paid for it pre-release, and there are 10 more books they have to work on to bring in more revenue instead. Real books have a permanence about them that motivates publishers to spend the extra time polishing their layout, tweaking the kerning, and all the other little things that Make a Difference to the overall reading experience.
Did I mention the fact that eReaders pretty much universally suck for non-sequential reading workflows? Kindle might be a net improvement over a 1,000 page printed novel, but it's not an improvement over a random printed Manning book. Every e-reader I've seen feels like you're wading through wet concrete when flipping through the pages.
And dammit, I'm still trying to find an Android pdf or epub reader that will let me read 2-up pages (like an open book) on my Xoom in landscape orientation (intelligently trimming away much of the margins, but maintaining the relative layout and flow of the printed book).
The other problem is that the hardware is just too underpowered. Books might be static while you're reading, but replicating the experience of flipping pages in a real book requires some major CPU power... not to mention fast, random-access media and lots of ram. None of this "fetch a tiny chunk at a time through a microSD single-bit cocktail straw" and "run at 200MHz to make the undersized battery last more than 2 hours" bullshit.
Verizon and Sprint had no choice but to upgrade (to LTE or Wimax), because EVDO rev.A hits a brick wall at ~2mbps. They COULD have upgraded to EVDO.revB, but the capital cost would have been almost the same, and they would have ended up locked in to Qualcomm as a single-source vendor forever, and paid premium prices for everything they bought going forward.
For AT&T and T-Mobile, the benefits of LTE aren't nearly as big. They're real, and they exist, but they aren't earth-shaking for 80% or so of their real-world users -- the 80% for whom HSPA+ already gives 16-20mbps, or for whom 1700MHz LTE would be unusable anyway. Remember, real-world HSPA is ~4-6mbps, and most HSPA+ is 10-16mbps. LTE might start at 16mbps and creep up to 26-30mbps (limited mainly by backhaul fiber capacity), but LTE's brick wall is signal strength -- below a certain point that's ~10dBm higher than the minimum needed by HSPA(+), LTE doesn't work *at all*.
I'm not 100% sure, but I believe that in a moving vehicle on a freeway (or a train), HSPA+ through urban areas might even work better and more consistently than LTE, simply because the phone can (theoretically) maintain at least one active tower connection at all times while leapfrogging between alternating streams. So... you might have tower 1 and 2 actively connected, lose tower 2, connect to tower 3, lose tower 1, connect to tower 4, lose tower 3, connect to tower 5, and so on... falling to 4-6mbps, bursting to 10-12mbps, and repeating over and over again, but always remaining connected by at least one or the other. In contrast, with LTE, you'd have periods of time with no connection at all when it broke the current connection to establish a new one with the next tower.
T-Mobile is deploying LTE, but in all honesty, the distinction between them is moot for probably 70-80% of their customers.
HSPA+, when it works properly, is basically as fast as LTE. The catch is, HSPA+ only gets its fastest rates if the user has solid connections to two or more towers, because it works the same way a dialup shotgun modem worked -- the phone independently connects to two or more towers, then splits the bitstream and sends part to each tower. Upstream, the bits are recombined into a single bitstream by T-Mobile's network.
Where LTE makes a difference is suburbia. Specifically, suburban locations where the user can only see one tower well, but has a rock-solid signal from that one tower. THEN, the user might get 16-20mbps from LTE, but only 4-6mbps from HSPA.
Here's the catch: if the user has a mediocre signal from two towers, but a strong signal from NONE, he might discover that T-Mobile's 1700MHz LTE doesn't work at all, and he's still limited to 10-12mbps HSPA+. If the user has a mediocre signal from ONE tower, he'll probably be lucky to see 1-2mbps, just like he does now, and LTE won't work either.
Why? LTE's throughput is kind of like 8-VSB TV transmissions. When it's strong enough, it's flawless and full-speed, even when other radio modes are degrading badly. But the moment your signal gets even a tiny bit weaker than the minimum (roughly 10dBm stronger than the minimum for viable HSPA), you fall off the cliff and lose it entirely. You don't get slower LTE... you have no LTE *whatsoever*.
So... one strong tower == LTE faster than HSPA(+)
Two mediocre towers == HSPA+ mostly works, LTE doesn't work reliably.
One mediocre tower == HSPA limps along, LTE doesn't work at all.
Two or more strong towers == LTE slightly faster than HSPA+ in theory, and might use less power, but looking at speed alone, you'd be hard-pressed to tell with any real certainty whether the user was using LTE or HSPA+. LTE has better latency (no need to demux, split, buffer, and recombine bitstreams, and the symbol rate itself is faster), but the difference isn't all that huge.
A 10kW generator is *barely* enough to run 2 or 3 10,000BTU window air conditioners. I believe you need ~24kW of 220v capacity to start a normal 2-3 ton central air conditioner.
It's a shame companies like Carrier, Rheem, etc can't put a little effort into designing central ac units that are "generator friendly" & can start with less inrush current. Like, maybe some kind of transmission that would allow the compressor to spin up slowly, instead of just soaking up 20+ kW for 3 seconds before settling down to half that amount. Or logic to start up the compressor, THEN the blower fan, instead of both at once.
Ummm... try testing that theory by tossing a chunk of elemental sodium or calcium into a bucket of water (or swimming pool, or jacuzzi). There are plenty of videos on Youtube illustrating the outcome...
AT&T (in my case, U-verse) might have an annoying cap, but it's big enough to not really bother me much. On the other had, my AT&T wireless cap (3 gigabytes) is terrifyingly low, because unlike the former cap, there's a real chance I could go over it and get charged extra in any given month. So I *do* actually care about the details of how they calculate it.
Specifically, is AT&T Wireless billing "chunky"? Suppose I have an Andriod app running in the background that relentlessly polls some remote server every 10 seconds. Or 60 seconds, if it matters. My hypothetical app tries to keep the bandwidth down, and works as follows:
User sends 1-byte command via UDP, then disconnects.
Server looks at the request to discern the sender's IP and port, decides what to send based on the command byte, and sends a one-byte response.
As far as I know, a UDP datagram with a single-byte payload is 33 bytes long (24 bytes for the IP header, 8 bytes for the UDP header, 1 byte for the payload). It's just a hunch, but I suspect that my likelihood of getting billed for exactly 66 bytes of data use is 100% pure fantasy. I'd be shocked if those two datagrams didn't end up getting billed as two 1k chunks of data, or worse.
Are wireless carriers required to file explicit tariffs with state utility regulators disclosing their exact charges, and explain how they calculate any usage-based fees in detail?
> Am I missing something here?
Yes. If the US has a shitload of shale, China and Russia almost certainly do, too & just haven't found it yet. And I think it's a bit premature to think their response will be to say, "This isn't Clean Green Energy(tm), so we mustn't be like those naughty Earth-hating Americans and use it."
Far MORE likely is that average European voters will get pissed off about higher energy prices in Europe, eventually elect leaders who aren't quite as determined to be Green at any cost, and join the celebration. And probably discover plenty of their own shale while they're at it.
Not even the most wildly-optimistic supporter of "green energy" fantasizes that it can produce as much cheap energy as fossil fuels and nuclear power. Their plans all involve "somehow" imposing life with less energy on the rest of the world.
In the long run, if you want to encourage the use of fuels that don't produce greenhouse gases, support nuclear power. It's far from perfect, and can be nasty stuff at times, but the truth is... it's the best thing we have, and the best thing we're likely to have for a really long time. If we reprocessed intensely-radioactive spent fuel instead of trying to store mass quantities of it on-site (with long-term plans to bury it under a mountain somewhere), the waste problem would mostly take care of itself.
If you think of spent nuclear fuel like Alkaline batteries, reprocessing is kind of like adding a "joule thief" circuit (which multiplies the voltage of a dying battery, even as it falls off the voltage cliff) and DC-DC regulator to a consumer electronics device that would normally just use 4 AA batteries to grab ~4.5 to 6 volts, so you can wring useful power out of the battery until it's literally squeezed dry and barely outputting enough DC to detect with a multimeter.
> (...snip...) We find some dirty old bag of crack (...snip...)
"We clear out the trash that's been down in the basement since we moved in, and discover an old, forgotten sub-basement built in the 19th century for coal that's filled with enough dirty old crack to to fill our own habit, plus the habits of the entire northeastern US, for the next 5000 years, and scream "Yeah! We're saved!..."
There, fixed that for you.
There are two things you never, ever, EVER want on your main webserver or app server: blog software (Wordpress, etc) and user-generated or uploaded content. Both are several orders of magnitude more likely to get your server pwn3d, and keeping them at arm's length from your main site & app server is kind of like building a ship with bulkhead doors to contain flooding in the event of a breach.
I speak from experience. Wordpress got totally owned somehow on its server. My app? Safe, 2500 miles away, on a server with no usernames/passwords/hostnames in common.
The problem is that apps like VBulletin, PhpBB, & others might actually be more secure than YOUR app, but your app doesn't have 47,000 hackers searching it for vulnerabilities & ready to exploit every site using the same software the moment a 0-day is found. Web forums are "high-value" targets. Your app, running one one server with a few thousand daily users, is not. You'll still get attacked, of course, but you're less likely to get compromised by some wacky, obscure bug in a system library that only affects a few servers (including yours), because THOSE exploits are only worth the time & effort of attackers when they can be automated or have enough potential victims to be worthwhile.
Another reason for a totally different hostname: to keep your whole site from getting blacklisted by Google over an XSS exploit some user pulled off in a posting.
The problem is that in the US, you (or a corporation) can't take a patent owner to court (seeking to overturn his patent) unless you/the corporation have "standing" -- ie, an infringement suit has been filed against you. And the patent's owner can drop the suit against you at any time up to the final moment the judge rules it invalid (preventing the troll from suing YOU again, but not others). I might be wrong, but I believe that in China, it *is* possible for a coalition of likely victims to gang up on the patent's owner & sue to get it overturned.
In the US, you CAN seek "re-examination", but it's expensive, the odds are overwhelmingly stacked against you, and (afaik), the only thing "winning" gets for you is immunity from treble damages if you act upon your victory, get sued for infringement anyway, and they prevail.
The huge bug in the US patent system is that patents are presumed valid unless overturned in court, but courts operate on the assumption that patents wouldn't have been granted without merit, and there have been plenty of instances (particularly the late 90s/early 00s) when overwhelmed examiners approved questionable patents on the theory that the courts could sort out their mess later.
The problem is compounded by American patent law's willingness to grant total blocking power to even the most incidental case of infringement. IE, if I hold the first patent on some primitive implementation that barely even works (if it works at all), and you *profoundly* improve upon it and make it viable, I can still block you from using it because my patent covers something more fundamental. I'm pretty sure that in China, there's a statutory process for compulsory licensing where someone who has the profoundly-improved patent could file the paperwork, then tell the owner of patent #1, "Shut up, take your fair share of the money, and file a lawsuit if you think you deserve more."
In the US, patent owners have too much power to prevent their patent's use by others at *any* cost, sane or otherwise.
There's no legal process to say, "Yes, I agree his patent is valid, but he's an asshole/troll/sociopath and is demanding more than its fair market value relative to the whole product, or is flat-out refusing to licence it at all." So, we have cases of companies restricting patent use to force you to go with one service provider or another -- RESTRICTING innovation, forcing consumers into false dichotomies, and preventing them from being able to have ideal devices that combine the best of everything -- from Edison's hated, inferior stock tickers to TiVo-vs-Dish/DirecTV/AT&T (cablecards don't work with them, so you CAN'T buy and use your own DVR) and Apple-vs-Samsung.
Moore's law says nothing about speed. Moore's law is about the number of transistors.
Netburst (a.k.a., the Pentium 4) was a miserable, failed experiment in CPU Marketecture (optimizing the wrong features just to give the marketing department big, meaningless numbers to dangle in front of consumers who don't know any better). It was designed from the start to offer cheap, worthless gigahertz, with less real performance than older CPUs running at half their nominal speed. May it rot in hell (and help keep it toasty and warm) forever.
In historical terms, the Pentium 4 family was kind of like trying to make freight trains faster by bolting a jet engine on top of a boxcar and using its exhaust to propel it. The Core family is metaporically like someone realizing they'd be better off using the jet engine to drive a turbine generator in the locomotive, buffer the power with a large bank of batteries, and use it to run bigger electric motors to push and pull the train from both ends instead.
Tell me about it. I have a nominally-1.5GHz quadcore Android phone that, when running Graffiti, can barely tell the difference between a "G" drawn like a "6" on the letter side, and the letter "O" with better than 90% accuracy unless I use SetCPU to lock it to full speed (with devastating impact upon battery life) whenever the screen is on, yet somehow... SOMEHOW... a slow, lowly 16MHz Dragonball m68k could do the same thing with nearly perfect, flawless accuracy. The biggest single reason, as far as I can tell? On the Palm, the digitizer was read via timer interrupt with clockwork regularity, and the recognition routine was 100% assembly. I think the Android version was written in Java, gets to have a sample whenever Android feels like allowing it to have one, and has its accuracy flushed down the toilet by endless speed-scaling that destroys the motion-vector algorithm that made the original so good, then made even worse by the insistence of Google (and Amazon, and everyone else) upon making AJAX network calls between every goddamn letter, with no real way to disable it (there used to be... briefly... but Google appears to have taken it away for good at some point between Gingerbread and ICS).
Sigh. Someday, I'm going to finally learn how to implement kernel-level Graffiti on Android (well, ok... kernel-level digitizer sampling with clockwork, non-scaled regularity that gets neatly buffered, so that the higher-level recognition routine can lurch and stumble as usual, but at least it'll have solid sample data to generate its motion vectors from).