Long ago I had a subscription to the System Internals newsletter. It was fascinating to peek behind the curtains, but as Linux matured, sadly I came to regard these excellent newsletters as episodes of reality TV. It was like peering into the details of a private life that ought to remain private, for the sanity of everyone involved.
Funny to contrast those two links by Russinovich. The first makes you think to yourself "abandon common sense, all those who enter here". In the second, he's saying "why is it that no-one preaches common sense about the size of your page file?" Eh?
Because too much of the Windows virtual memory architecture is a ritual of waving dead chickens. An ordinary person can't keep straight in his mind which parts require a dead chicken and which parts don't, so we hew to our mimetic talismans.
At one point he is running a contrived process to exhaust his system's commit limit, and the process explorer view of memory doesn't manage to point the finger at the flagrant transgression. So when Windows configures my pagefile limit to three times my 4GB system memory, or whatever notion it gets, I feel like it's sitting there daring me to shrink it: feel lucky, punk?
I've known forever that it's a stupid formula relative to what a pagefile *ought* to be doing. But I don't know it is *actually* doing what it ought to be doing, unless I wish to wade into encyclopedia Russinovich, which is entertaining, but not worth a long term commit charge. For all I know, some badly coded device driver requires the pagefile to be three times larger than Walmart, for a reason no sane OS could possibly dream up.
I also wonder about the strange neglect of the orphan memory over 4G in Windows 32.
I understand you can't map this memory into any user process on speaking terms with a (badly coded) device driver, but why the heck can't this otherwise dormant pool of memory be used as a preferred chunk of RAM based pagefile?
The OS kernel has complete control of page loading. It could certainly load the page as an above 4G to below 4G page copy, rather than an expensive disk operation. No badly coded device driver discombobulated as a result. Heck, a clever Northbridge might even do this on your behalf.
Some of this is valid technical limits. Some of this is a limit of corporate will, bamboozled behind technical limits.
Having said all that, I don't understand the original complaint.
I run XP at work with a pretty heavy workload. I have a desktop manager with nine desktops. I have a start script that reloads most of my key applications into their assigned desktops. (Unfortunately, I haven't figured out how to get the Firefox session manager to restore windows to their previous desktops.) It takes about ten minutes for my system to settle down after a reboot. But after that, the delays associated with switching into a long idle desktop have never struck me as unbearable.
When I have seen terrible application switching delays, it was usually on some system geared for video games, under far less non-game load than my development system at work (which is a very ordinary machine).
Another huge culprit I've seen is background indexers which thrash the file cache. On my gf's laptop, all of the nasty delays were eventually traced to a WiFi network browser. Then there are the geniuses that run some high-end OpenGL screen saver with massive textures, who then wonder why it's all a bit sluggish when the screen saver exits.
It's pretty rare I've seen a system where the pagefile itself was to blame for XP slowness. My experience is that XP doesn't page out excessively unless something else is pressuring it to do so. If you then disable the pagefile, that something else will be forced to run more slowly, while your main apps will remain responsive. If that something else is a passel of viri, or a posse of anti-viri, then I can see how disabling the pagefile might be perceived as progress.
Nope. 1GB is 16x the minimum required for XP Home and XP Pro, which require a minimum of 64MB.
That's like defining "minimum" tire tread on your car by what it takes to keep your muffler from scraping the asphalt. If you jimmied the leaf springs, I bet XP Home would boot in 32M and play a mean game of FreeCell.
Good UI is not about making a UI that learns the user - a computer will never be able to do a good job of that.
I'm going to write that one down beside the meme leader: that no matter how much hard drive capacity you have, the next version of the Microsoft OS will expand to fill it. This was the universal anguished cry of the mid to late 1990s.
Update for 2010: I hear a rumor that Windows 7 will ship with over a billion distinct holographic avatars, so that no two customers share the same image. They've managed to fit this into 1.997 TB, so a Windows 7 install on a 2 TB disk drive will only leave you with 3 GB of free space.
The difference between saying that "no adaptive algorithm will ever be any good" and "man will never land on the moon" is that the later was once defensible.
Adaptive interfaces has more to do with human psychology than intrinsic difficulty with machine learning (which is not to say the problem is yet solved).
Humans are used to dealing with other humans, and we assume a highly contextualized interaction. Computers have never been given access to a context as rich as we assume in our most casual human interactions. When we program computers for adaptive behaviour without providing this context, the result strikes us as oscillating wildly between inspired and moronic.
If a software module managed to obtain as much context as a living human would obtain another person in the same room would be regarded as pretty darn invasive. Wouldn't you just love to see the singular value decomposition of the perceptive android that shares your living space?
on the rag 70% on outs with boyfriend 30% bloating due to tofu lunch 45% neediness 85% competitive grooming 79% criticism rejection 33% sarcastic infighting 61%
We perceive little progress in machine learning because we have such a myopic gold standard. If computers made any real progress toward our myopic gold standard, we'd be scared as hell.
Furthermore, the incentives are dubious. What's the rate limiting step in personal productivity? Mousemanship? Man, if only my mouse were perfectly tuned to my nervous system, I could have rendered the whole of Ratatouille in under a day.
How many discussion threads does one need to read to figure out that for most of the human race, rising to the level of processing ideas would be a mental stretch. Most of what goes on in the human brain under the guise of thinking is bickering and rhetoric dressed up as insight. Perhaps the rate limiting factor on human productivity is how we process information, not how we push it around? In which case, how much economic incentive is there to optimize the flow of pixels to the n'th adaptive degree?
Another problem in making an adaptive desktop is that the end user is not a pure game theoretic construct. Generally the user experience is a two headed monster: one head is conventionally labeled "consumer desire" while the other head is labeled "corporate overlord".
The two heads have competing interests, which corporate overlord is busy downplaying until consumer desire wearily acquiesces. Relative to consumer desire, a truly adaptive DVD player would skip past those obnoxious FBI warnings.
You can't make an appliance adaptive to consumer desire "except when" it conflicts with corporate overlord, because that would require corporate overlord to define the boundary of the control imposed. Not going to happen. So we get quasi adaptation, within narrow parameters, which isn't adaptation at all, and Coke vs Pepsi, or two old white guys on the next ballot posing as choice.
Should we be careful what we wish for? What is our true appetite for adaptive interfaces? How much should it pander to our bad habits? How many times should an adaptive desktop optimize the interface to enable you to more quickly select individual files using a GUI picker before it reaches out and c
I've done ARM7 development in the past. I've just started a new ARM7 project after a four year hiatus. What seems to have changed the most is the availability of cheap hardware debug tools such as cheap FTDI FT2232 based JTAG pods.
I'm also interested to see what comes of Eclipse's DSDP initiative.
I've always thought the Balkanization of the debug hardware was one of the major barriers in the wider circulation of open hardware.
Concerning the Adweebo, I've programmed this chip before using the free download (for Windows) AVR Studio. What an amazing tool. Despite the AVR having a dedicated hardware call stack, this thing can't even display a call stack.
Nor can it display local variables for call frames up the stack. The upshot of this is that if you have a complicated protocol subroutine that calls get_byte() which blocks waiting for the next byte of input from USART or SPI/TWI (as slave), then whenever you break the program (yes, you must break the program for AVR Studio to display *anything*) you'll inevitably end up looking at the local variables of get_byte() which won't be interesting, while all the variables you wish to inspect in the calling routine are unwatchable.
In another nod to genius, whenever you break execution under AVR Studio, it changes focus to the source code tab where the execution happens to break. Even if you just had your cursor on the variable you wished to watch, or the line of code where you wished to add a break point, neither of which can be done while the program is running. Breaking the program changes your view, and then you have to find it again among twenty or so tabs you might have open.
I've never managed to develop much proficiency with GDB. I expect my new project will finally cure that. Generally I write my code so that I don't spend time debugging at that level.
It's nice to see prebuilt packages such as Yagarto where the GDB to random JTAG-of-the-month debug interface is pre-configured, and you're not forced to invest $700 in a "professional" level debug pod to get debug features that we really ought to take for granted in this day and age.
Also, I can't sign off without coughing up a hairball to describe the modalities of the Atmel debugWire interface. Bluuurp. There, that feels better.
Reading this thread, I finally get the origins of the expression CYA: it's the Kevlar pants some people wear to protect innocent bystanders against the sonic boom when the the large carbon briquette they're clenching suddenly compresses into that small diamond of professional livelihood.
In a bookstore the other day, I was leafing through Zittrain's "The Future of the Internet--And How to Stop It". I didn't take much from it myself, but it's a fine summary of many issues to pass along to that clueful someone outside the IT profession.
"The many rules strip us of the most important thing: the ability to be considerate. We're losing our capacity for socially responsible behavior," says Dutch traffic guru Hans Monderman, one of the project's co-founders. "The greater the number of prescriptions, the more people's sense of personal responsibility dwindles."...
Psychologists have long revealed the senselessness of such exaggerated regulation....
The plans derive inspiration and motivation from a large-scale experiment in the town of Drachten in the Netherlands, which has 45,000 inhabitants. There, cars have already been driving over red natural stone for years. Cyclists dutifully raise their arm when they want to make a turn, and drivers communicate by hand signs, nods and waving.
"More than half of our signs have already been scrapped," says traffic planner Koop Kerkstra. "Only two out of our original 18 traffic light crossings are left, and we've converted them to roundabouts." Now traffic is regulated by only two rules in Drachten: "Yield to the right" and "Get in someone's way and you'll be towed."
Strange as it may seem, the number of accidents has declined dramatically.
So much for the rigid linear relationship between rule conformance and orderly outcomes so beloved among the pride-in-earning-to-eat-another-day crowd.
That said, I think eggs are juvenile outlets, and utterly inappropriate for any class of software that can result in a more serious injury than a bent nose.
At the same time, I don't feel professionalism demands such a cold relationship with your work that your emotional investment never rises above rule conformance and personal CYA blast containment shields.
And no, don't think I'm advocating Verkeersbordvrij as a universal cure-all. I'm not even convinced its initial success will hold up.
I'd imagine that with a little cleverness and effort, it'd be possible to come up with a rain jacket design that had a decent amount of venting in places that were adequately protected from rain.
Or you could shop for an hour in the pre-fall season at any outdoor clothing store in Vancouver, BC, if your definition of "rain" is every shade of grey from mist to drizzle.
As a practical matter "out of box" is synonymous with lowest common denominator.
The dominant criteria in tuning OOB is that the OS be tolerable for almost any conceivable workload within the capability of the machine.
If you are also able to tweak performance for commonplace tasks (such as the LAMP stack) *without* compromising the former goal, you can do that too, to some degree. Bear in mind not to turn on any performance features that are poorly documented, have surprising edge cases, or security considerations that aren't widely understood.
Let me try again in visual terms. OOB is what results when you ship a Ferrari engine to Volvo who then dress it up as an eight passenger station wagon with air bags, headrests, and seatbelts for every occupant.
Try to picture it. Two long columns of eight-seat station wagons packed together at the start line of the Volvo Indianapolis 500 rumbling their engines. Some powered by Solaris, some by Ubuntu engines, some by FreeBSD.
This is actually a profound subject, though few of the posts I've read seem to agree. The youthful fear age, so they mock it.
What I recall reading is that functional memory remains strong in our middle years, but we tend to lose incidental memories, memories about other things we were doing at the time.
In my work, we've just designed a circuit board around a processor family I used once before, in a contract five years ago. I recall the architecture quite well. I recall which tasks caused me more pain than expected. But I don't recall the exact names of the tools I used. And I used these tools daily for more than a year. What was the name of that crappy debugger and that stupid JTAG pod? Ten years ago, chances I would have forgotten these details were essentially zilch. Once I did a little research on the web, it all came back to me, but I had to trigger the recollections externally.
I suspect the availability of memory is one of the main reasons many mathematicians and scientists do their best work at a young age. In order to notice obscure lateral connections, you need powerful recall. Matching a pattern is a form of memory.
Even with my memory in noticeable decline (which galls me, but what can you do?) I hardly feel obsolete. It's like I'll never run the marathon in 2:30 again. Does that make me unfit? Hardly, but some of the people around me might now remember more than I do. At this stage in my career I won't be the first to put up my hand to take on the task of programming to an API of 200 badly designed, irregular functions. (Unless it's PHP. When you have an API of 2000 irregular functions, it puts us all on the same footing. We're all programming with on hand on the PHP guide book. But I digress.)
In the age of Google, it's not clear to me what we actually *need* to remember. One thing I decided is that before I get really old, I'm going to train myself to check a memory aid before repeating the same question out loud over and over because I can't remember that I've already asked the question. Unfortunately, by the time you can't remember that you've asked a question, you're too far gone to train this habit. I'll think I'll put a note on my day timer for my eightieth birthday reminding me to get on this before it's too late.
And the other side of the coin is that IT is not a producing industry.
You sound like a old-fashioned taxonomist who doesn't want to admit that Tyrannosaurus tastes like chicken.
Are you trying to tell me that DNA is more important than mitochondia? That you could rip out the space shuttle's electronic nervous system and it would become "less efficient"? That brain death is merely a quality of life issue?
I think it's a good thing if we succeed in this race to the bottom spearheaded by the gift community of open source. For the next fifty years or so, the "more with more" pendulum is going to be slapping us in the face.
Reminds me of an anecdote about some famous scientist whose biography I once read.
"Mommy, what if all the good problems are solved before I grow up?"
"Don't worry, pet, I'm sure there's plenty of problems to go around."
Everyone here who looks into the future and sees a less complicated world, please raise your hands.
The fact of the matter is that any country whose citizens are in need of food aid has a governing class who feels even more greatly deprived. It's an unsolved problem how to distribute food without paying the tin pot tax.
The only way to be certain about where your aid dollar ends up at the end of the day is to distribute machine guys or lingerie.
RISC machines made sense before Intel figured out to make x86 go faster than one instruction per clock.
That's a good starting point, but you missed half the story.
The concept that RISC failed to stir into their soup is that latency is fundamentally asynchronous in general purpose computing. There are specialized floating-point kernels and such where latency can be successfully regimented into a synchronous model. The majority of customers with these requirements bought dedicated, specialized machines. It kept a few of these specialized designs alive, but didn't leave enough on the table to fund a $2 billion fab.
What made the Pentium Pro insanely great is that it accomplished as much for bandwidth and latency as it did for throughput. Aggregate bandwidth to the L1, L2, and main memory was shocking to their competitors. Not measured as peak bandwidth, but as sustained bandwidth on unruly general-purpose code mixes.
I read many discussions where the Alpha camp sneered at Intel for resorting to the expensive on chip (but off die) 256KB L2 cache. They seemed to forget their own imprudent expenses, such as an entire extra layer of expensive metal interconnect to handle carry propagation in their single-cycle 64 bit adders. I might not remember all this exactly right, it's been a very long time.
In my opinion, every chip that's bet against asynchronous latency has stumbled. Itanium was a bet on regimenting latency, as can be judged by its performance (plenty) on the (relatively rare) tasks where latency isn't much of an issue. Many RISC chips grind to a appalling halt when an L1 cache request misses.
What asynch latency means is that you never know for certain what order you will execute the upcoming instruction stream. That's a larger proportion of your grey hair than the RISC/CISC distinction.
It was always the case that the memory subsystem was chewing up die area faster than the execution core. Once the core frequency to EBI multiple exceeds 3:1, the core itself is no longer your largest design headache.
For the x86, the often disparaged RMW instruction format goes a long way toward decreasing bandwidth pressure on the L1 cache interface. Almost enough to make up for the paltry physical register set.
Where I bet the 3000 person Intel design team cursed their forebears was in the area of partial register stalls and the irregular flag register. When every instruction modifies a different subset of your flag register bits, it must be a nightmare to track this in an OOO processor.
Thumb2 pretty much proves that mixed instruction lengths was the right approach from the outset. It was stupid of x86 to permit arbitrary combinations of prefix bytes and instruction lengths varying anywhere from 1 to 16 bytes. Two (or maybe three) well judged instruction lengths goes a long way toward achieving a happy balance of code density, i-cache density, i-cache bandwidth, and throughput.
I've always been amused that the RISC camp brags about their instruction decoders using so many hundred-thousand transistors less than CISC, while their i-cache code density on their multi-megabyte i-caches is 30% lower. Well, it sure improved time to market that you just cut and paste more i-cache rather than design convoluted instruction decoders. Shame about the capacitance of the i-cache drive lines. Biting into the frequency budget? Nothing another layer of custom metal interconnect won't solve. It's not as if we're selling so many that fabrication time will become a bottleneck... uh, wait a minute, let's back up and rethink that.
It's pretty clear that the Achilles heal of x86 was heat, not performance. The problem is that Intel's team of 3000 designers actually succeeded in keeping most of those transistors busy most of the time.
Intel's later blindness to this problem baffles me. You have to wonder if some Intel execs held a lot of Enron stock, back when it was still worth something. If 5 mill
Surprisingly, the left side of the keyboard actually includes more than half the letters of the alphabet...
I wonder if this guy had a former job in finance. He sounds like Greenspan on the shocking and unprecedented and totally unanticipated breakdown of counterparty surveillance.
I'm at an age where I'm not sure if I'm for or against the idea that I've passed the mid-point in my life expectancy. Having some engineering training himself, my father taught me binary around the time the 4004 was introduced, with an cardboard egg carton and some marbles.
He also borrowed on my behalf *all three* of the terrible "modern era" computer books from the local university library. One of these books focused on photographs of IBM consoles there were already headed toward obsolescence when the IBM Selectric was the epitome of modernism.
One of the books had a tolerable explanation of boolean logic.
The third book was all about the use of flow charts to document and express algorithms. Since "home" computers didn't entirely exist yet, I tried drawing flow charts to help me get to school on time. It never entered my mind that perhaps the problem had something to do with staying up until 1:00am every school night reading any book I could get my hands on.
By the time the local school was teaching us how to multiply a pair of two digit numbers (a proficiency one gains in binary much sooner), I had determined to my own satisfaction that flow charts were a misguided tool, at best, in the depiction of intellectual property.
For the purposes of this discussion, I have the distinction of being on the downslope of life, while computers remain as much "in my blood" as a recent college grad who "discovered" computers as a young child before Linux had its first GUI. I'm of two minds on this question.
What's the extra two decades worth? Most of the time, I suspect not all that much. I wonder about this meme of older people having learned from their past mistakes. You read about that in Dilbert sometimes. People do learn, yes, but rarely in a good way.
Certainly, one gets better at blaming other people. Erecting taller fences around job responsibilities. Not landing the undoable core component on your own plate. Not panicking as much under deadline pressure, because you knew all along that the young guy next to you has a component twice as impossible as your own.
The ultimate accomplishment of middle-aged conservatism is to finish an aggressive project right on time, right on budget, and have the finished circuit not actually able to perform a real-life task of any commercial utility. I've seen the baby discarded with the bath water in service to "delivered on time, paid on time".
Learning from the past is an extremely variable skill. Some people have none of it, some people have a fair amount, and some people do or don't with a shocking mixture of randomness.
Prudence is usually accompanied by lessening of enthusiasm. Is it a good exchange? Not always.
I think I did my best work 15 years ago. Never on time, but I moved the rock in a way I rarely even attempt these days.
Compared to younger co-workers, my analytic and written skills are for the most part vastly superior.
My ability to pastiche together an almost-working prototype our of inferior code "just lying around" is not so good. I never mastered the trick of "almost working". Somewhere along the line I acquired the personal baggage that there is no reason an algorithm or a core block of code shouldn't be 100% correct. This is baggage I now struggle with daily. Older people have baggage. I might have more than my share.
Lately I'm working on an embedded project. I download sample code from the chip vendor. I look at the code and go "ugh, this was not developed with a spirit of purity and elegance". Then I freeze up like a deer in the headlights. Younger people don't seem to have this problem. From what I call tell, most younger people have been pastiching crap together since the day they first learned to type. Crap is situation normal, a core proficiency.
After curling up in a ball and muttering "architecture" to myself a few dozen times, I come out of my trance, and manage to find a way to move forward again, without rewriting gobs of ucky code to my
A quick search came up with one site listing the cost of Samarium as $130 per 100g. I'm sure that's cost effective for medical imaging equipment. And I had never realized this, but our local landfill is positively brimming with discarded medical scanning equipment. I might try to scavenge some of this, but all the discarded MRI machines are clumped together by some unseen force.
Does watching your 401k melt into nothingness count as unhappiness?
The flip side question is more interesting. Does your 401K climb into the stratosphere on the basis of money that never existed in the first place count as happiness?
Most of the recent "performance" of the markets was banks enlarging the money supply with complex financial instruments while risks disappeared under the sofa cushion.
What's more, many of these fund managers were collecting advance fees, and massing their excessive compensation in offshore bank accounts (we can only presume).
So what is the relationship between gullibility and happiness, after all?
Why is it that technocratic hair-splitters distinguish every point of law except the difference between law and fear of law?
Packet inspection for the purpose of resource management hardly goes against the spirit of common carrier. Does anyone really think you could pack a dozen shot guns into the back of stage coach, and not have the stage coach driver know what goods you were dealing in? Do you suppose no gold miner ever was told by his local stage coach operator, "we're not too keen on porting your two bundles of dynamite; maybe you should seek transport by other means?"
This might have nothing to do with the actual law concerning common carrier status. I'm just pointing out that it's quite ridiculous to think any common carrier ever has been 100% ignorant of the cargo they convey. Even first class mail is sometimes signed hugs and kisses and botulism.
Fortunately, the judges sometimes manage to make jurisprudence work, despite the text of the law being often stacked against them. Which is not to say that judges are free to interpret legislation they dislike differently than the legislation is written. But they can resort to generic precedent to put forward the view the legislation is too preposterous to enforce in 100% literalness.
A good example of preposterous legislation would be a bill passed in the era of stagecoaches being construed to mean that AOL and their like can be put to the thumbscrews by the powerful political lobby of content owners. If that's what congress wants, they should update the law to unambiguously say as much.
Since we don't know what a judge would actually do, we're not talking so much about law, as fear of law. Deploying deep packet inspection opens an ISP up to the claim, by deeper pockets, that they are no longer operating within the safe confines of common carrier.
Fear of the law is the regime in effect when people reason "we would probably prevail if we could see this through, but we don't have enough money to find out".
Fear of law boils down either to lazy government, or bad government bought and paid for. We shouldn't have to sit around trembling over how some stage-coach era legislation would be construed by a higher courts, if a deep pocket stepped forward to contest the matter.
Really, the legislative and judiciary systems need to get together once in a while for a clean-out-the-fridge party where the dozen eggs from 1850 are finally put to rest. The way it seems to work is that old law just keeps getting pushed further to the back of the fridge, no matter how musty and inedible, it doesn't get thrown out.
The legal profession has done a pretty good job of creating a generalized fear of opening the fridge. Somehow, despite our purported democracy, we allow them to get away with this. It makes no more sense to me than paying a banker $500m to bankrupt the powerful corporation he works for. Part of the problem is that too many among us exult in fear of law and extolling its minutia.
the whole premise is cracked
on
Googling Security
·
· Score: 4, Interesting
The premise here is "if only we had known ahead of time, we would have done things differently". In the cases where we did know ahead of time, or enough people did, we still went ahead and did it anyway. *After* the Grand Banks fishery collapsed... we continued to fish it. A few short years later... we shut down the entire fishery due to lack of foresight and cooperation.
For some reason, I've never viewed Google as a particularly large threat. They seem to be using the data mining to sell a well targeted audience. Is there a Google service where I can pay to get dirt on my neighbors? There's two guys living out front I'd like to get rid of.
Like a bank, there is a business model to make a lot of money in a hurry by whisking all the deposits off to an island paradise. However, the business model where they maintaining the trust relationship with the fools who deposited in the first place pays better in the long run. When you get down to it, banks sell trust, and not much else.
Do we think our banks don't know a lot about us? If only we had known, we'd have never allowed banks to exist in the first place.
What's happening here is that with mass storage plummeting into the $/TB range, one way or another we were going to have to rethink our entire privacy and public information models rather dramatically.
If only we had known, we'd have never allowed Shugart to spin that first platter.
Actually, yes I will, because McCain's losing ratio was less than his "coverage" ratio.
That pretty much maxes out my disingenuous crap detector.
Over at 538 the other day, they did a regression on voter contact in the ground game. I think the result was that for a 10% margin in direct voter contact, Obama outperformed his poll by 3%. First of all, these aren't linear relations, although there might be regions of quasi-linearity. Secondly, even if there is a linear region, the slope coefficient does not default to 1:1, except in the flattest of all flat earth conceptions. You're exaggeration media influence by a factor to 10 to 100 times.
But you think you can get away with this bollocks exaggeration because you've chose the round number "1" as your regression coefficient, despite the fact that not a single electoral parameter has a 1:1 coefficient other than votes for candidate to ballots marked for candidate, and even that one is often suspect.
Speaking of stupid metrics, I'd be curious to see the result of a Palin:Biden reportage analysis. In my view, every article about Palin also counts as an article on McCain's judgment, whether his name was mentioned or not.
I had an inkling to print a bumper sticker
Land of the brave: vote Palin.
I'm reminding of the saying normally attributed to Woody Allen: "90% of Life is Showing Up"
There seems to be a conception that our media must conform to this view by definition: each presidential candidate gets equal coverage by virtue of showing up with a warm pulse, unless your name is Ralph Nader (I guess that's the other 10%). It doesn't matter what you say or do, you deserve equal column inches.
This whole equity notion is cracked. Even the "warm pulse" criteria favored Obama. I dread to think of the barrels of ink spilled if McCain had been elected the day his warm pulse failed to materialize.
It's not as if the average voter can't expose themselves to any ratio of column inches they might desire, except expats based in Europe and the Middle East, but that isn't the media they were analyzing in this stupid report.
I don't know how you can blame the media in Europe, either. The headline "Bush is a dipshit" only holds the reader's attention the first 500 times. Almost as tiresome as Clinton's pickle, which got more coverage than both of these candidates combined.
Speaker wire is the reason "unlimited" will never exist in pure form. The same people who purchase $8,000 speaker wires are quite convinced that even if they were capped at 1TB/hour for their holographic porn, it would still be a curly hair shy of the real thing.
I'd have no problem with capped download if the cap decayed at a sensible exponential rate, the same way that gmail's free storage ticks ever upward. If the cap doubled every two years (corresponding to a 40% annual cost reduction in the cost of carrying traffic, which I'm certain the optical portions of the backbone achieve), then ten years from now, the current monthly cap would have evolved into the daily cap. At that rate, you're already watching a three hour HD movie every day of your life, or multibooting every Linux distro that every existed at the same time onto your 256 core processor.
Depending on the cost of your speaker wire, this might or might not suffice.
This from a major book vendor who can't manage to position the ISBN number within the first screen of a book's product page and which feeds up a biased selection of user reviews without a great deal of extra clicking.
Which of my frustrations were addressed when Gillette added a battery to a razor which already has more blades than I can bother to count? Strangely, I can't recall.
Also, one must admire the way Gillette integrated a consumable which changes the blade angle to something less optimal even if the blades fail to promptly become dull.
Of course, if the packaging weren't such a bitch, I'd be happy as a pig in mud to pitch my razor head into the landfill every second shave.
The hypothetical particle even seems to have the right mass to account for one theory of dark matter."
I'm so disappointed. I thought you were going to ask the usefully blatant question: would it have been possible to discover a particle with a mass that didn't fit at least one theory of dark matter? If the stupid thing had weighed a kilogram, there is probably some (totally cracked) theory of dark matter out there it would fit into perfectly. The problem with modern physics is that "theory of everything" turned into "theory for everything" with a parameter space of 2^500, to bandy around another number consistent with something a physicist somewhere recently scratched onto a blackboard.
Re:Mebbe I should try it some time
on
OpenBSD 4.4 Released
·
· Score: 4, Insightful
How is it that these comments are raised again and again with rarely a genuflect towards the possibility that our social norms and our technical norms exist at cross purposes?
It is often pointed out that humans are hierarchical animals. What's pointed out far less often is that we are also polarizing animals. For the most part, it's pretty darn hard to get a community of people to rest comfortably within a dual hierarchy: the polarizers will either succeed in driving the culture toward a political hierarchy, or they will succeed in driving the culture toward a technical meritocracy, politics be damned.
What evidence do we have that people can be effective and polite at the same time? NASA? I think not. When it became a political culture, shuttles exploded.
Is Linus an ass, or does he choose to occupy the niche that has proven viable? Larry Wall has taken a gentler stance toward his position as benevolent dictator for life, and he's not getting much good press lately. Nice guys finish last or at best, five years late.
Every time this subject comes up, there is a lot of chattering from the "How to win friends and influence people" crowd that despite the technical merits of X, it doesn't suit that person's social worldview, as if technical merit belongs in a marriage with popularity and approval.
As far as I can tell from my experience, the majority of PC marriages of that ilk are functionally destitute, yet the chattering never ceases that the world *ought* to operate that way. On what basis? What annoys me most is that this chattering rarely includes even the slightest nod toward justification.
This is another fact about human nature: we seem to have an inbuilt algorithm for determining that certain kinds of opinions can be safely put forward with little or no justification (e.g. "that's just how things are"), and which kinds of opinion can automatically be called to account. In my experience, the hierarchy of what must be fully justified and what needn't be has been pretty much decided on the grade 3 playground.
There seems to be a lot of people out there who are offended to the core that Theo's objectionable personality has been associated with so much durable accomplishment. In my opinion, that's just a bad case of shooting the messenger. Given broad human instincts toward hierarchy and polarization, it was as inevitable as the rise of the spam king having created a zero-cost anonymous distribution channel.
The underlying problem is that there is no reliable chalk line between civility and brown-nosing, and it's hell to police in a project that could otherwise rely on more objective measures. It's kind of like Sudoku. A complete waste of time, but I enjoy it anyway. We've made almost no progress (as a social organism) at efficiently policing the line between civility and brown-nosing, but so many among our ranks seem to prefer sliding down this slippery moss bank over the firm traction of dystopian merit.
I'd be interested to hear what danger this small lump of metal would pose to human health after (barely) surviving incineration at a temperature sufficient to vaporize steel. Isn't that kind of like warning people to steer clear of any metal instrument that has just come out of an autoclave?
I suspect the agenda has more to do with manufacturing a general fear in the mind of the public toward any debris that falls from the sky. I've always wondered whether hydrazine was an essential design element of an F16, or merely gets included for its scarecrow factor. It seems any space technology the Americans wish to keep secret either contains hydrazine or something intensely radioactive.
That certainly makes for a better public service announcement than "don't go near anything that falls from the sky because of the tens of thousands of dollars many foreign governments would pay for a covert photograph of what you've inadvertently discovered".
Nobody planned out this current hideous jumble of technologies we call teh intertubez,
What makes you say that? Not all forms of planning are central planning. I think the plan was to create a network where diversity could flourish, and solve problems that actually occurred, rather than solving problems *before* they actually occurred.
The biggest mistake was the lag in crytographic protocols concerning email and DNS which would have helped to combat spam and spoofing. One needs to recall that crytography was mired in a political and patent minefield in the time period when these technologies could have been proven out, before the intertubes were energized. It was U.S. government policy to delay the global proliferation of strong crypto, and they succeeded for about a decade or so.
The internet has already scaled six orders of magnitude, and nothing goes horribly wrong on a day to day basis. I've never heard Google say "if only the internet had been designed better, we might have succeeded".
The only way to claim "they got it wrong" is to have completely forgotten, in retrospect, the number of ways it could have gone much worse.
The Intel CPU only scaled over three orders of magnitude over frequency (4MHz to 4GHz) before things started to go sideways (multicore). Really, you need a clear head to think up anything that has scaled better than the internet, and in less time, given where it began.
Good grief, if you had a time machine, would you go back twenty years to tweak the designers of TCP/IP on the nose, or would you go back thirty years to slap Intel upside the head concerning the design of the 8086? I say that knowing the 8086 has succeeded far beyond the level that any of its original detractors would have imagined possible.
If the segment register had been an eight bit offset instead of a four bit offset, the machine would have had a 16MB address space, every expansion card could have had a dedicated 64KB address range. This would have nicely covered the period of time before VM actually starts to work well, around 16MB of physical RAM.
If I could only send one twitter back in time (in the greybeard advice category), that would be it.
Of course, that would have enlarged the market for memory, potentially at the expense of how much people paid for their CPU, which might not have made Intel quite as successful as they were.
The technoverse sure looks funny where you're an entrepreneur attempting to monopolize the micro-content niche halfway between a punch line and a six o'clock news sound bite. Wish I could trademark white space. I'd like to have 25 cents for everyone out there who would pay twitter to go away.
Nothing stops the distromakers from making their distro boot in 5 seconds.
That's one the least informed comments I've seen in a long time.
If you haven't noticed this before, protocols such as TCP have built-in time scales which don't change according to the speed of your processor. A modern hard drive can only perform about 100 random seeks per second. If booting up requires more than 500 random seeks, you aren't going to achieve a 5s boot time. It's pretty darn easy for a file system to require 500 seeks to load many dozens of independent device drivers, although the original FAT file system would save you quite a bit by not having any sub-directories to traverse.
And it's not as if all the daemons and applications only look in one place for their configuration files. It's a small miracle if X-windows can load its default fonts in less than 500 disk seeks.
Between the BIOS interfaces, probing for devices that don't exist, loading device drivers that do exist, the daemons, and the windowing system, you might have 50 to 100 independent upstream maintainers for the components activated during the boot process.
And then, of course, you have to be able to log and report errors that might occur in any step along the chain.
There's also a lot of things that have to start from scratch so as not to potentially interfere with the environment around you. You can't just assume you can send packets on the same IP address, netmask, and gateway you were assigned on last boot: the network configuration might have changed in the meantime, your leased address might have been reassigned, etc.
Not to mention the boot process navigates thirty years worth of backwards compatibility cruft. Seen a version of glibc that compiles to less than 1MB lately? Do you think glibc really *needs* to be that bloated? Or does it have more to do with being compatible with 10,000 source packages?
But I agree with you. If we tipped 30 years of hardware design and legacy software into the landfill, we could boot in under 5s, no sweat.
Slashdot Mirror
Long ago I had a subscription to the System Internals newsletter. It was fascinating to peek behind the curtains, but as Linux matured, sadly I came to regard these excellent newsletters as episodes of reality TV. It was like peering into the details of a private life that ought to remain private, for the sanity of everyone involved.
Funny to contrast those two links by Russinovich. The first makes you think to yourself "abandon common sense, all those who enter here". In the second, he's saying "why is it that no-one preaches common sense about the size of your page file?" Eh?
Because too much of the Windows virtual memory architecture is a ritual of waving dead chickens. An ordinary person can't keep straight in his mind which parts require a dead chicken and which parts don't, so we hew to our mimetic talismans.
At one point he is running a contrived process to exhaust his system's commit limit, and the process explorer view of memory doesn't manage to point the finger at the flagrant transgression. So when Windows configures my pagefile limit to three times my 4GB system memory, or whatever notion it gets, I feel like it's sitting there daring me to shrink it: feel lucky, punk?
I've known forever that it's a stupid formula relative to what a pagefile *ought* to be doing. But I don't know it is *actually* doing what it ought to be doing, unless I wish to wade into encyclopedia Russinovich, which is entertaining, but not worth a long term commit charge. For all I know, some badly coded device driver requires the pagefile to be three times larger than Walmart, for a reason no sane OS could possibly dream up.
I also wonder about the strange neglect of the orphan memory over 4G in Windows 32.
I understand you can't map this memory into any user process on speaking terms with a (badly coded) device driver, but why the heck can't this otherwise dormant pool of memory be used as a preferred chunk of RAM based pagefile?
The OS kernel has complete control of page loading. It could certainly load the page as an above 4G to below 4G page copy, rather than an expensive disk operation. No badly coded device driver discombobulated as a result. Heck, a clever Northbridge might even do this on your behalf.
Some of this is valid technical limits. Some of this is a limit of corporate will, bamboozled behind technical limits.
Having said all that, I don't understand the original complaint.
I run XP at work with a pretty heavy workload. I have a desktop manager with nine desktops. I have a start script that reloads most of my key applications into their assigned desktops. (Unfortunately, I haven't figured out how to get the Firefox session manager to restore windows to their previous desktops.) It takes about ten minutes for my system to settle down after a reboot. But after that, the delays associated with switching into a long idle desktop have never struck me as unbearable.
When I have seen terrible application switching delays, it was usually on some system geared for video games, under far less non-game load than my development system at work (which is a very ordinary machine).
Another huge culprit I've seen is background indexers which thrash the file cache. On my gf's laptop, all of the nasty delays were eventually traced to a WiFi network browser. Then there are the geniuses that run some high-end OpenGL screen saver with massive textures, who then wonder why it's all a bit sluggish when the screen saver exits.
It's pretty rare I've seen a system where the pagefile itself was to blame for XP slowness. My experience is that XP doesn't page out excessively unless something else is pressuring it to do so. If you then disable the pagefile, that something else will be forced to run more slowly, while your main apps will remain responsive. If that something else is a passel of viri, or a posse of anti-viri, then I can see how disabling the pagefile might be perceived as progress.
That's like defining "minimum" tire tread on your car by what it takes to keep your muffler from scraping the asphalt. If you jimmied the leaf springs, I bet XP Home would boot in 32M and play a mean game of FreeCell.
I'm going to write that one down beside the meme leader: that no matter how much hard drive capacity you have, the next version of the Microsoft OS will expand to fill it. This was the universal anguished cry of the mid to late 1990s.
Update for 2010: I hear a rumor that Windows 7 will ship with over a billion distinct holographic avatars, so that no two customers share the same image. They've managed to fit this into 1.997 TB, so a Windows 7 install on a 2 TB disk drive will only leave you with 3 GB of free space.
The difference between saying that "no adaptive algorithm will ever be any good" and "man will never land on the moon" is that the later was once defensible.
Adaptive interfaces has more to do with human psychology than intrinsic difficulty with machine learning (which is not to say the problem is yet solved).
Humans are used to dealing with other humans, and we assume a highly contextualized interaction. Computers have never been given access to a context as rich as we assume in our most casual human interactions. When we program computers for adaptive behaviour without providing this context, the result strikes us as oscillating wildly between inspired and moronic.
If a software module managed to obtain as much context as a living human would obtain another person in the same room would be regarded as pretty darn invasive. Wouldn't you just love to see the singular value decomposition of the perceptive android that shares your living space?
on the rag 70%
on outs with boyfriend 30%
bloating due to tofu lunch 45%
neediness 85%
competitive grooming 79%
criticism rejection 33%
sarcastic infighting 61%
We perceive little progress in machine learning because we have such a myopic gold standard. If computers made any real progress toward our myopic gold standard, we'd be scared as hell.
Furthermore, the incentives are dubious. What's the rate limiting step in personal productivity? Mousemanship? Man, if only my mouse were perfectly tuned to my nervous system, I could have rendered the whole of Ratatouille in under a day.
How many discussion threads does one need to read to figure out that for most of the human race, rising to the level of processing ideas would be a mental stretch. Most of what goes on in the human brain under the guise of thinking is bickering and rhetoric dressed up as insight. Perhaps the rate limiting factor on human productivity is how we process information, not how we push it around? In which case, how much economic incentive is there to optimize the flow of pixels to the n'th adaptive degree?
Another problem in making an adaptive desktop is that the end user is not a pure game theoretic construct. Generally the user experience is a two headed monster: one head is conventionally labeled "consumer desire" while the other head is labeled "corporate overlord".
The two heads have competing interests, which corporate overlord is busy downplaying until consumer desire wearily acquiesces. Relative to consumer desire, a truly adaptive DVD player would skip past those obnoxious FBI warnings.
You can't make an appliance adaptive to consumer desire "except when" it conflicts with corporate overlord, because that would require corporate overlord to define the boundary of the control imposed. Not going to happen. So we get quasi adaptation, within narrow parameters, which isn't adaptation at all, and Coke vs Pepsi, or two old white guys on the next ballot posing as choice.
Should we be careful what we wish for? What is our true appetite for adaptive interfaces? How much should it pander to our bad habits? How many times should an adaptive desktop optimize the interface to enable you to more quickly select individual files using a GUI picker before it reaches out and c
I've done ARM7 development in the past. I've just started a new ARM7 project after a four year hiatus. What seems to have changed the most is the availability of cheap hardware debug tools such as cheap FTDI FT2232 based JTAG pods.
I'm also interested to see what comes of Eclipse's DSDP initiative.
http://www.eclipse.org/dsdp/dsdp-charter.php
I've always thought the Balkanization of the debug hardware was one of the major barriers in the wider circulation of open hardware.
Concerning the Adweebo, I've programmed this chip before using the free download (for Windows) AVR Studio. What an amazing tool. Despite the AVR having a dedicated hardware call stack, this thing can't even display a call stack.
Nor can it display local variables for call frames up the stack. The upshot of this is that if you have a complicated protocol subroutine that calls get_byte() which blocks waiting for the next byte of input from USART or SPI/TWI (as slave), then whenever you break the program (yes, you must break the program for AVR Studio to display *anything*) you'll inevitably end up looking at the local variables of get_byte() which won't be interesting, while all the variables you wish to inspect in the calling routine are unwatchable.
In another nod to genius, whenever you break execution under AVR Studio, it changes focus to the source code tab where the execution happens to break. Even if you just had your cursor on the variable you wished to watch, or the line of code where you wished to add a break point, neither of which can be done while the program is running. Breaking the program changes your view, and then you have to find it again among twenty or so tabs you might have open.
I've never managed to develop much proficiency with GDB. I expect my new project will finally cure that. Generally I write my code so that I don't spend time debugging at that level.
It's nice to see prebuilt packages such as Yagarto where the GDB to random JTAG-of-the-month debug interface is pre-configured, and you're not forced to invest $700 in a "professional" level debug pod to get debug features that we really ought to take for granted in this day and age.
Also, I can't sign off without coughing up a hairball to describe the modalities of the Atmel debugWire interface. Bluuurp. There, that feels better.
Reading this thread, I finally get the origins of the expression CYA: it's the Kevlar pants some people wear to protect innocent bystanders against the sonic boom when the the large carbon briquette they're clenching suddenly compresses into that small diamond of professional livelihood.
In a bookstore the other day, I was leafing through Zittrain's "The Future of the Internet--And How to Stop It". I didn't take much from it myself, but it's a fine summary of many issues to pass along to that clueful someone outside the IT profession.
Had a nice passage concerning Verkeersbordvrij.
From http://www.spiegel.de/international/spiegel/0,1518,448747,00.html
So much for the rigid linear relationship between rule conformance and orderly outcomes so beloved among the pride-in-earning-to-eat-another-day crowd.
That said, I think eggs are juvenile outlets, and utterly inappropriate for any class of software that can result in a more serious injury than a bent nose.
At the same time, I don't feel professionalism demands such a cold relationship with your work that your emotional investment never rises above rule conformance and personal CYA blast containment shields.
And no, don't think I'm advocating Verkeersbordvrij as a universal cure-all. I'm not even convinced its initial success will hold up.
Or you could shop for an hour in the pre-fall season at any outdoor clothing store in Vancouver, BC, if your definition of "rain" is every shade of grey from mist to drizzle.
I couldn't agree less.
As a practical matter "out of box" is synonymous with lowest common denominator.
The dominant criteria in tuning OOB is that the OS be tolerable for almost any conceivable workload within the capability of the machine.
If you are also able to tweak performance for commonplace tasks (such as the LAMP stack) *without* compromising the former goal, you can do that too, to some degree. Bear in mind not to turn on any performance features that are poorly documented, have surprising edge cases, or security considerations that aren't widely understood.
Let me try again in visual terms. OOB is what results when you ship a Ferrari engine to Volvo who then dress it up as an eight passenger station wagon with air bags, headrests, and seatbelts for every occupant.
Try to picture it. Two long columns of eight-seat station wagons packed together at the start line of the Volvo Indianapolis 500 rumbling their engines. Some powered by Solaris, some by Ubuntu engines, some by FreeBSD.
Enjoy the race, buddy.
This is actually a profound subject, though few of the posts I've read seem to agree. The youthful fear age, so they mock it.
What I recall reading is that functional memory remains strong in our middle years, but we tend to lose incidental memories, memories about other things we were doing at the time.
In my work, we've just designed a circuit board around a processor family I used once before, in a contract five years ago. I recall the architecture quite well. I recall which tasks caused me more pain than expected. But I don't recall the exact names of the tools I used. And I used these tools daily for more than a year. What was the name of that crappy debugger and that stupid JTAG pod? Ten years ago, chances I would have forgotten these details were essentially zilch. Once I did a little research on the web, it all came back to me, but I had to trigger the recollections externally.
I suspect the availability of memory is one of the main reasons many mathematicians and scientists do their best work at a young age. In order to notice obscure lateral connections, you need powerful recall. Matching a pattern is a form of memory.
Even with my memory in noticeable decline (which galls me, but what can you do?) I hardly feel obsolete. It's like I'll never run the marathon in 2:30 again. Does that make me unfit? Hardly, but some of the people around me might now remember more than I do. At this stage in my career I won't be the first to put up my hand to take on the task of programming to an API of 200 badly designed, irregular functions. (Unless it's PHP. When you have an API of 2000 irregular functions, it puts us all on the same footing. We're all programming with on hand on the PHP guide book. But I digress.)
In the age of Google, it's not clear to me what we actually *need* to remember. One thing I decided is that before I get really old, I'm going to train myself to check a memory aid before repeating the same question out loud over and over because I can't remember that I've already asked the question. Unfortunately, by the time you can't remember that you've asked a question, you're too far gone to train this habit. I'll think I'll put a note on my day timer for my eightieth birthday reminding me to get on this before it's too late.
You sound like a old-fashioned taxonomist who doesn't want to admit that Tyrannosaurus tastes like chicken.
Are you trying to tell me that DNA is more important than mitochondia? That you could rip out the space shuttle's electronic nervous system and it would become "less efficient"? That brain death is merely a quality of life issue?
I think it's a good thing if we succeed in this race to the bottom spearheaded by the gift community of open source. For the next fifty years or so, the "more with more" pendulum is going to be slapping us in the face.
Reminds me of an anecdote about some famous scientist whose biography I once read.
"Mommy, what if all the good problems are solved before I grow up?"
"Don't worry, pet, I'm sure there's plenty of problems to go around."
Everyone here who looks into the future and sees a less complicated world, please raise your hands.
Starting with the word "we" this screams moronic.
The fact of the matter is that any country whose citizens are in need of food aid has a governing class who feels even more greatly deprived. It's an unsolved problem how to distribute food without paying the tin pot tax.
The only way to be certain about where your aid dollar ends up at the end of the day is to distribute machine guys or lingerie.
That's a good starting point, but you missed half the story.
The concept that RISC failed to stir into their soup is that latency is fundamentally asynchronous in general purpose computing. There are specialized floating-point kernels and such where latency can be successfully regimented into a synchronous model. The majority of customers with these requirements bought dedicated, specialized machines. It kept a few of these specialized designs alive, but didn't leave enough on the table to fund a $2 billion fab.
What made the Pentium Pro insanely great is that it accomplished as much for bandwidth and latency as it did for throughput. Aggregate bandwidth to the L1, L2, and main memory was shocking to their competitors. Not measured as peak bandwidth, but as sustained bandwidth on unruly general-purpose code mixes.
I read many discussions where the Alpha camp sneered at Intel for resorting to the expensive on chip (but off die) 256KB L2 cache. They seemed to forget their own imprudent expenses, such as an entire extra layer of expensive metal interconnect to handle carry propagation in their single-cycle 64 bit adders. I might not remember all this exactly right, it's been a very long time.
In my opinion, every chip that's bet against asynchronous latency has stumbled. Itanium was a bet on regimenting latency, as can be judged by its performance (plenty) on the (relatively rare) tasks where latency isn't much of an issue. Many RISC chips grind to a appalling halt when an L1 cache request misses.
What asynch latency means is that you never know for certain what order you will execute the upcoming instruction stream. That's a larger proportion of your grey hair than the RISC/CISC distinction.
It was always the case that the memory subsystem was chewing up die area faster than the execution core. Once the core frequency to EBI multiple exceeds 3:1, the core itself is no longer your largest design headache.
For the x86, the often disparaged RMW instruction format goes a long way toward decreasing bandwidth pressure on the L1 cache interface. Almost enough to make up for the paltry physical register set.
Where I bet the 3000 person Intel design team cursed their forebears was in the area of partial register stalls and the irregular flag register. When every instruction modifies a different subset of your flag register bits, it must be a nightmare to track this in an OOO processor.
Thumb2 pretty much proves that mixed instruction lengths was the right approach from the outset. It was stupid of x86 to permit arbitrary combinations of prefix bytes and instruction lengths varying anywhere from 1 to 16 bytes. Two (or maybe three) well judged instruction lengths goes a long way toward achieving a happy balance of code density, i-cache density, i-cache bandwidth, and throughput.
I've always been amused that the RISC camp brags about their instruction decoders using so many hundred-thousand transistors less than CISC, while their i-cache code density on their multi-megabyte i-caches is 30% lower. Well, it sure improved time to market that you just cut and paste more i-cache rather than design convoluted instruction decoders. Shame about the capacitance of the i-cache drive lines. Biting into the frequency budget? Nothing another layer of custom metal interconnect won't solve. It's not as if we're selling so many that fabrication time will become a bottleneck ... uh, wait a minute, let's back up and rethink that.
It's pretty clear that the Achilles heal of x86 was heat, not performance. The problem is that Intel's team of 3000 designers actually succeeded in keeping most of those transistors busy most of the time.
Intel's later blindness to this problem baffles me. You have to wonder if some Intel execs held a lot of Enron stock, back when it was still worth something. If 5 mill
I wonder if this guy had a former job in finance. He sounds like Greenspan on the shocking and unprecedented and totally unanticipated breakdown of counterparty surveillance.
I'm at an age where I'm not sure if I'm for or against the idea that I've passed the mid-point in my life expectancy. Having some engineering training himself, my father taught me binary around the time the 4004 was introduced, with an cardboard egg carton and some marbles.
He also borrowed on my behalf *all three* of the terrible "modern era" computer books from the local university library. One of these books focused on photographs of IBM consoles there were already headed toward obsolescence when the IBM Selectric was the epitome of modernism.
One of the books had a tolerable explanation of boolean logic.
The third book was all about the use of flow charts to document and express algorithms. Since "home" computers didn't entirely exist yet, I tried drawing flow charts to help me get to school on time. It never entered my mind that perhaps the problem had something to do with staying up until 1:00am every school night reading any book I could get my hands on.
By the time the local school was teaching us how to multiply a pair of two digit numbers (a proficiency one gains in binary much sooner), I had determined to my own satisfaction that flow charts were a misguided tool, at best, in the depiction of intellectual property.
For the purposes of this discussion, I have the distinction of being on the downslope of life, while computers remain as much "in my blood" as a recent college grad who "discovered" computers as a young child before Linux had its first GUI. I'm of two minds on this question.
What's the extra two decades worth? Most of the time, I suspect not all that much. I wonder about this meme of older people having learned from their past mistakes. You read about that in Dilbert sometimes. People do learn, yes, but rarely in a good way.
Certainly, one gets better at blaming other people. Erecting taller fences around job responsibilities. Not landing the undoable core component on your own plate. Not panicking as much under deadline pressure, because you knew all along that the young guy next to you has a component twice as impossible as your own.
The ultimate accomplishment of middle-aged conservatism is to finish an aggressive project right on time, right on budget, and have the finished circuit not actually able to perform a real-life task of any commercial utility. I've seen the baby discarded with the bath water in service to "delivered on time, paid on time".
Learning from the past is an extremely variable skill. Some people have none of it, some people have a fair amount, and some people do or don't with a shocking mixture of randomness.
Prudence is usually accompanied by lessening of enthusiasm. Is it a good exchange? Not always.
I think I did my best work 15 years ago. Never on time, but I moved the rock in a way I rarely even attempt these days.
Compared to younger co-workers, my analytic and written skills are for the most part vastly superior.
My ability to pastiche together an almost-working prototype our of inferior code "just lying around" is not so good. I never mastered the trick of "almost working". Somewhere along the line I acquired the personal baggage that there is no reason an algorithm or a core block of code shouldn't be 100% correct. This is baggage I now struggle with daily. Older people have baggage. I might have more than my share.
Lately I'm working on an embedded project. I download sample code from the chip vendor. I look at the code and go "ugh, this was not developed with a spirit of purity and elegance". Then I freeze up like a deer in the headlights. Younger people don't seem to have this problem. From what I call tell, most younger people have been pastiching crap together since the day they first learned to type. Crap is situation normal, a core proficiency.
After curling up in a ball and muttering "architecture" to myself a few dozen times, I come out of my trance, and manage to find a way to move forward again, without rewriting gobs of ucky code to my
A quick search came up with one site listing the cost of Samarium as $130 per 100g. I'm sure that's cost effective for medical imaging equipment. And I had never realized this, but our local landfill is positively brimming with discarded medical scanning equipment. I might try to scavenge some of this, but all the discarded MRI machines are clumped together by some unseen force.
The flip side question is more interesting. Does your 401K climb into the stratosphere on the basis of money that never existed in the first place count as happiness?
Most of the recent "performance" of the markets was banks enlarging the money supply with complex financial instruments while risks disappeared under the sofa cushion.
What's more, many of these fund managers were collecting advance fees, and massing their excessive compensation in offshore bank accounts (we can only presume).
So what is the relationship between gullibility and happiness, after all?
Why is it that technocratic hair-splitters distinguish every point of law except the difference between law and fear of law?
Packet inspection for the purpose of resource management hardly goes against the spirit of common carrier. Does anyone really think you could pack a dozen shot guns into the back of stage coach, and not have the stage coach driver know what goods you were dealing in? Do you suppose no gold miner ever was told by his local stage coach operator, "we're not too keen on porting your two bundles of dynamite; maybe you should seek transport by other means?"
This might have nothing to do with the actual law concerning common carrier status. I'm just pointing out that it's quite ridiculous to think any common carrier ever has been 100% ignorant of the cargo they convey. Even first class mail is sometimes signed hugs and kisses and botulism.
Fortunately, the judges sometimes manage to make jurisprudence work, despite the text of the law being often stacked against them. Which is not to say that judges are free to interpret legislation they dislike differently than the legislation is written. But they can resort to generic precedent to put forward the view the legislation is too preposterous to enforce in 100% literalness.
A good example of preposterous legislation would be a bill passed in the era of stagecoaches being construed to mean that AOL and their like can be put to the thumbscrews by the powerful political lobby of content owners. If that's what congress wants, they should update the law to unambiguously say as much.
Since we don't know what a judge would actually do, we're not talking so much about law, as fear of law. Deploying deep packet inspection opens an ISP up to the claim, by deeper pockets, that they are no longer operating within the safe confines of common carrier.
Fear of the law is the regime in effect when people reason "we would probably prevail if we could see this through, but we don't have enough money to find out".
Fear of law boils down either to lazy government, or bad government bought and paid for. We shouldn't have to sit around trembling over how some stage-coach era legislation would be construed by a higher courts, if a deep pocket stepped forward to contest the matter.
Really, the legislative and judiciary systems need to get together once in a while for a clean-out-the-fridge party where the dozen eggs from 1850 are finally put to rest. The way it seems to work is that old law just keeps getting pushed further to the back of the fridge, no matter how musty and inedible, it doesn't get thrown out.
The legal profession has done a pretty good job of creating a generalized fear of opening the fridge. Somehow, despite our purported democracy, we allow them to get away with this. It makes no more sense to me than paying a banker $500m to bankrupt the powerful corporation he works for. Part of the problem is that too many among us exult in fear of law and extolling its minutia.
The premise here is "if only we had known ahead of time, we would have done things differently". In the cases where we did know ahead of time, or enough people did, we still went ahead and did it anyway. *After* the Grand Banks fishery collapsed ... we continued to fish it. A few short years later ... we shut down the entire fishery due to lack of foresight and cooperation.
For some reason, I've never viewed Google as a particularly large threat. They seem to be using the data mining to sell a well targeted audience. Is there a Google service where I can pay to get dirt on my neighbors? There's two guys living out front I'd like to get rid of.
Like a bank, there is a business model to make a lot of money in a hurry by whisking all the deposits off to an island paradise. However, the business model where they maintaining the trust relationship with the fools who deposited in the first place pays better in the long run. When you get down to it, banks sell trust, and not much else.
Do we think our banks don't know a lot about us? If only we had known, we'd have never allowed banks to exist in the first place.
What's happening here is that with mass storage plummeting into the $/TB range, one way or another we were going to have to rethink our entire privacy and public information models rather dramatically.
If only we had known, we'd have never allowed Shugart to spin that first platter.
That pretty much maxes out my disingenuous crap detector.
Over at 538 the other day, they did a regression on voter contact in the ground game. I think the result was that for a 10% margin in direct voter contact, Obama outperformed his poll by 3%. First of all, these aren't linear relations, although there might be regions of quasi-linearity. Secondly, even if there is a linear region, the slope coefficient does not default to 1:1, except in the flattest of all flat earth conceptions. You're exaggeration media influence by a factor to 10 to 100 times.
But you think you can get away with this bollocks exaggeration because you've chose the round number "1" as your regression coefficient, despite the fact that not a single electoral parameter has a 1:1 coefficient other than votes for candidate to ballots marked for candidate, and even that one is often suspect.
Speaking of stupid metrics, I'd be curious to see the result of a Palin:Biden reportage analysis. In my view, every article about Palin also counts as an article on McCain's judgment, whether his name was mentioned or not.
I had an inkling to print a bumper sticker
Land of the brave: vote Palin.
I'm reminding of the saying normally attributed to Woody Allen: "90% of Life is Showing Up"
There seems to be a conception that our media must conform to this view by definition: each presidential candidate gets equal coverage by virtue of showing up with a warm pulse, unless your name is Ralph Nader (I guess that's the other 10%). It doesn't matter what you say or do, you deserve equal column inches.
This whole equity notion is cracked. Even the "warm pulse" criteria favored Obama. I dread to think of the barrels of ink spilled if McCain had been elected the day his warm pulse failed to materialize.
It's not as if the average voter can't expose themselves to any ratio of column inches they might desire, except expats based in Europe and the Middle East, but that isn't the media they were analyzing in this stupid report.
I don't know how you can blame the media in Europe, either. The headline "Bush is a dipshit" only holds the reader's attention the first 500 times. Almost as tiresome as Clinton's pickle, which got more coverage than both of these candidates combined.
Speaker wire is the reason "unlimited" will never exist in pure form. The same people who purchase $8,000 speaker wires are quite convinced that even if they were capped at 1TB/hour for their holographic porn, it would still be a curly hair shy of the real thing.
I'd have no problem with capped download if the cap decayed at a sensible exponential rate, the same way that gmail's free storage ticks ever upward. If the cap doubled every two years (corresponding to a 40% annual cost reduction in the cost of carrying traffic, which I'm certain the optical portions of the backbone achieve), then ten years from now, the current monthly cap would have evolved into the daily cap. At that rate, you're already watching a three hour HD movie every day of your life, or multibooting every Linux distro that every existed at the same time onto your 256 core processor.
Depending on the cost of your speaker wire, this might or might not suffice.
This from a major book vendor who can't manage to position the ISBN number within the first screen of a book's product page and which feeds up a biased selection of user reviews without a great deal of extra clicking.
Which of my frustrations were addressed when Gillette added a battery to a razor which already has more blades than I can bother to count? Strangely, I can't recall.
Also, one must admire the way Gillette integrated a consumable which changes the blade angle to something less optimal even if the blades fail to promptly become dull.
Of course, if the packaging weren't such a bitch, I'd be happy as a pig in mud to pitch my razor head into the landfill every second shave.
I'm so disappointed. I thought you were going to ask the usefully blatant question: would it have been possible to discover a particle with a mass that didn't fit at least one theory of dark matter? If the stupid thing had weighed a kilogram, there is probably some (totally cracked) theory of dark matter out there it would fit into perfectly. The problem with modern physics is that "theory of everything" turned into "theory for everything" with a parameter space of 2^500, to bandy around another number consistent with something a physicist somewhere recently scratched onto a blackboard.
How is it that these comments are raised again and again with rarely a genuflect towards the possibility that our social norms and our technical norms exist at cross purposes?
It is often pointed out that humans are hierarchical animals. What's pointed out far less often is that we are also polarizing animals. For the most part, it's pretty darn hard to get a community of people to rest comfortably within a dual hierarchy: the polarizers will either succeed in driving the culture toward a political hierarchy, or they will succeed in driving the culture toward a technical meritocracy, politics be damned.
What evidence do we have that people can be effective and polite at the same time? NASA? I think not. When it became a political culture, shuttles exploded.
Is Linus an ass, or does he choose to occupy the niche that has proven viable? Larry Wall has taken a gentler stance toward his position as benevolent dictator for life, and he's not getting much good press lately. Nice guys finish last or at best, five years late.
Every time this subject comes up, there is a lot of chattering from the "How to win friends and influence people" crowd that despite the technical merits of X, it doesn't suit that person's social worldview, as if technical merit belongs in a marriage with popularity and approval.
As far as I can tell from my experience, the majority of PC marriages of that ilk are functionally destitute, yet the chattering never ceases that the world *ought* to operate that way. On what basis? What annoys me most is that this chattering rarely includes even the slightest nod toward justification.
This is another fact about human nature: we seem to have an inbuilt algorithm for determining that certain kinds of opinions can be safely put forward with little or no justification (e.g. "that's just how things are"), and which kinds of opinion can automatically be called to account. In my experience, the hierarchy of what must be fully justified and what needn't be has been pretty much decided on the grade 3 playground.
There seems to be a lot of people out there who are offended to the core that Theo's objectionable personality has been associated with so much durable accomplishment. In my opinion, that's just a bad case of shooting the messenger. Given broad human instincts toward hierarchy and polarization, it was as inevitable as the rise of the spam king having created a zero-cost anonymous distribution channel.
The underlying problem is that there is no reliable chalk line between civility and brown-nosing, and it's hell to police in a project that could otherwise rely on more objective measures. It's kind of like Sudoku. A complete waste of time, but I enjoy it anyway. We've made almost no progress (as a social organism) at efficiently policing the line between civility and brown-nosing, but so many among our ranks seem to prefer sliding down this slippery moss bank over the firm traction of dystopian merit.
I'd be interested to hear what danger this small lump of metal would pose to human health after (barely) surviving incineration at a temperature sufficient to vaporize steel. Isn't that kind of like warning people to steer clear of any metal instrument that has just come out of an autoclave?
I suspect the agenda has more to do with manufacturing a general fear in the mind of the public toward any debris that falls from the sky. I've always wondered whether hydrazine was an essential design element of an F16, or merely gets included for its scarecrow factor. It seems any space technology the Americans wish to keep secret either contains hydrazine or something intensely radioactive.
That certainly makes for a better public service announcement than "don't go near anything that falls from the sky because of the tens of thousands of dollars many foreign governments would pay for a covert photograph of what you've inadvertently discovered".
What makes you say that? Not all forms of planning are central planning. I think the plan was to create a network where diversity could flourish, and solve problems that actually occurred, rather than solving problems *before* they actually occurred.
The biggest mistake was the lag in crytographic protocols concerning email and DNS which would have helped to combat spam and spoofing. One needs to recall that crytography was mired in a political and patent minefield in the time period when these technologies could have been proven out, before the intertubes were energized. It was U.S. government policy to delay the global proliferation of strong crypto, and they succeeded for about a decade or so.
The internet has already scaled six orders of magnitude, and nothing goes horribly wrong on a day to day basis. I've never heard Google say "if only the internet had been designed better, we might have succeeded".
The only way to claim "they got it wrong" is to have completely forgotten, in retrospect, the number of ways it could have gone much worse.
The Intel CPU only scaled over three orders of magnitude over frequency (4MHz to 4GHz) before things started to go sideways (multicore). Really, you need a clear head to think up anything that has scaled better than the internet, and in less time, given where it began.
Good grief, if you had a time machine, would you go back twenty years to tweak the designers of TCP/IP on the nose, or would you go back thirty years to slap Intel upside the head concerning the design of the 8086? I say that knowing the 8086 has succeeded far beyond the level that any of its original detractors would have imagined possible.
If the segment register had been an eight bit offset instead of a four bit offset, the machine would have had a 16MB address space, every expansion card could have had a dedicated 64KB address range. This would have nicely covered the period of time before VM actually starts to work well, around 16MB of physical RAM.
If I could only send one twitter back in time (in the greybeard advice category), that would be it.
Of course, that would have enlarged the market for memory, potentially at the expense of how much people paid for their CPU, which might not have made Intel quite as successful as they were.
The technoverse sure looks funny where you're an entrepreneur attempting to monopolize the micro-content niche halfway between a punch line and a six o'clock news sound bite. Wish I could trademark white space. I'd like to have 25 cents for everyone out there who would pay twitter to go away.
That's one the least informed comments I've seen in a long time.
If you haven't noticed this before, protocols such as TCP have built-in time scales which don't change according to the speed of your processor. A modern hard drive can only perform about 100 random seeks per second. If booting up requires more than 500 random seeks, you aren't going to achieve a 5s boot time. It's pretty darn easy for a file system to require 500 seeks to load many dozens of independent device drivers, although the original FAT file system would save you quite a bit by not having any sub-directories to traverse.
And it's not as if all the daemons and applications only look in one place for their configuration files. It's a small miracle if X-windows can load its default fonts in less than 500 disk seeks.
Between the BIOS interfaces, probing for devices that don't exist, loading device drivers that do exist, the daemons, and the windowing system, you might have 50 to 100 independent upstream maintainers for the components activated during the boot process.
And then, of course, you have to be able to log and report errors that might occur in any step along the chain.
There's also a lot of things that have to start from scratch so as not to potentially interfere with the environment around you. You can't just assume you can send packets on the same IP address, netmask, and gateway you were assigned on last boot: the network configuration might have changed in the meantime, your leased address might have been reassigned, etc.
Not to mention the boot process navigates thirty years worth of backwards compatibility cruft. Seen a version of glibc that compiles to less than 1MB lately? Do you think glibc really *needs* to be that bloated? Or does it have more to do with being compatible with 10,000 source packages?
But I agree with you. If we tipped 30 years of hardware design and legacy software into the landfill, we could boot in under 5s, no sweat.