"It's interesting to read how you so clearly hate the idea of someone else having control, and at the same time you want that exact same control all to yourself above the devs. That's the most common problem with the middle-management level of development, they're in charge and they're going to have it their way, all other factors be damned."
I AM a dev, you idiot. Back when I was fresh and new and guileless, like you, I thought a lot like you. Everything I knew about development had come from two sources, CS assignments and Slashdot--I didn't even know how little I knew.
Most of that got that beat out of me, eventually--not by management, but by other engineers who just didn't have the patience for my snotty, immature attitude.
"What the development team as a whole needs is balance. I'm not saying for either side to hold things hostage, I'm just saying don't let the PHB's decide when it's "ready" based solely on a trade show date."
And now you're just redefining your side of the argument because I burned it out from under you, in my previous comment. Changing your tune, now, cannot erase your previous comments. YHL, HAND.
"...athough you gave a reasonable example of how there is only a finite range of shoelace lengths that are practical..."
NO. That's NOT what I did.
I demonstrated that:
* While there is a THEORETICALLY INFINITE RANGE of possible theoretical shoelace lengths;
* Nonetheless, the probability distribution of actual shoelace lengths assigns a non-infinitessimal (and rather large) probability to a finite (and rather small) range of lengths.
Do you remember WHY that's important? It's because in your previous post, YOU CLAIMED that:
"...in an infinite field of possibilities, everything is infinitesimally rare."
Which is just utter horseshit. You denied the existence of uneven probability distributions--seriously, what the fuck is wrong with you?
You are now, officially, squirming like a fish on a hook. You have been owned at every comment, by pretty much everybody, probably because your math training is well below average for Slashdot.
You can keep trying to sneak your way out of this argument, so you can maintain your fragile sense of self-worth, but nothing will change the fact that you got your ass royally whooped.
Go back to school, kiddo, and pay more attention in class.
So... You're a perfectionist who will hold a project hostage until YOU decide, on your terms, that it's not "crap"? And anybody who disagrees with your assessment of costs versus benefits is a "pinhead"?
Remind me to hire you, really. I love programmers who sneer at business concerns, refuse to compromise with their teams, and are so inexperienced that they believe that any business can get done that way.
Man, if I had a whole team of guys like you, I'd be able to... Hmm. People like you actually detract from the productivity of a team. With enough of your clones running around bragging about how big their dicks are, business would grind to a halt. Hell, with a whole company of you, we could probably put global software productivity into reverse, it would be so bad.
"...if I had a really large project..."
I think this says it, right here, folks. If you ever you DO get on a large project, give me a call. Maybe you'll have learned enough to hire you.
Who said we were talking about MS and Windows? You just brought that up, right now. I don't think it proves anything, one way or another, that one company has a crappy process.
Honestly, it seems like you just tried to "move the goalposts", redefining the terms of an argument you were losing so you can feel like you're winning.
You incorrectly assume that "an infinite number of different programs" and "all possible programs" are the same set. They are not.
Turing's proof shows that no algorithm can solve the Halting Problem for *all possible* programs. But there ARE proven algorithms that solve the the Halting Problem for certain classes of programs, that is, subsets of "all possible" programs.
Many of those subsets (all the interesting ones, really) containg an infinite number of possible programs. Not *all* possible programs, mind you--just an infinitely-large subset of another distinct, infinitely-large superset.
(Please read the Wikipedia page on infinity, and grok the part about different sizes of infinity, before we continue. Don't worry, I've got time.
Got it? Good.)
Also, you wrote "...in an infinite field of possibilities, everything is infinitesimally rare." This statement is mathematically absurd--obviously, you've never taken probability, so you don't know how to talk about it in a useful way. Let me demonstrate:
There are an infinite number of possible lengths--the entire real number space, right? But not all shoelace lengths are all equally probable in that space. I'd guess that more than 99% of all shoelaces are shorter than 1 meter--my local drug store doesn't stock anything longer than about 75 cm.
Now, does that mean that my drugstore's policy "can't work"? Of course not--it works fine, they make a profit. If a clown with giant shoes needs mile-long shoelaces, nobody cares. It's perfectly useful to only solve 99% of the possible cases, right?
Back to programs: IF you define the probability distribution of human programmer output as totally random, you'd have a point. But humans don't bang out random strings of gibberish machine instructions like monkeys at typewriters. We tend to follow certain well-worn paths, and some of these subsets have proven Halting Problem-solvable. Tools that implement those insights aren't "general algorithms", but they're still useful.
Go read up on tools like Electric Fence, and other automated debugging aides. This is not a new concept, and you should really learn more about the subjects (math, software). It's embarrassing, for you.
How large of a programming team do you work with? And how big are the projects to which you contribute code? And what kind of development model do you use (waterfall, Agile, ad-hoc, etc.)?
Shipping a large project with 1,000 bugs might be a perfectly valid decision. Are any of those 1,000 bugs deal-breakers for your install base? If so, how many clients does it affect? Are these "real bugs", or just incomplete/unpolished functions, or documentation issues, or output typos, or what?
And what kind of software is this? Are you building a time & expense web application, or a filesystem driver? In the former case, most bugs will be interface glitches--ugly, annoying, and harmless. In the latter case, even one bug could easily cause silent data corruption.
Remeber what Linus Torvalds said: Release early, release often. Don't wait til all your bugs are fixed before shipping your software, or you'll lose your "market" window. If it's good enough, the early-adopters will understand, and might even contribute bug reports or patches that will speed you up.
Has anybody every told you "'Perfect' is the enemy of 'good enough'."? Perhaps after listening to you explain why your project is behind schedule, then sighing and face-palming?
The halting problem says that there cannot be a GENERAL ALGORITHM that works in all cases, for any of the infinity of possible programs that can exist.
That proves ZERO about, say, whether I can write an algorithm that covers 99% of the common cases. The lack of a general solution doesn't imply that it can't be done often enough, in practice.
The simplest scheme in real estate fraud is to rent a house, fill out some fake land title transfer paperwork, and then impersonate the lawful owner and either (A) sell the house to an unsuspecting, honest buyer, or (B) take out a bank mortgage on the house. Either way, you walk off with cash-in-hand.
Read and learn: http://www.abanet.org/genpractice/newsletter/lawtrends/0709/realestate/realestatefraud101.html . Google for many more news stories.
The really nasty part, here, is that the original homeowner is, in legal terms, cold fucked. In the case of an outright sale, the unsuspecting buyer actually has a stronger title claim to the property than the original owner--fraud notwithstanding. If the crook mortgaged the house, that bank has a possible (though weaker) legal claim on seizing the collateral (house) if nobody makes the payments.
The fundamental legal problem: In the US and Canada, the criminal offense of fraud is a totally separate matter from the civil property issue (the title claims, or the mortgage). A subsequent civil transaction isn't automatically voided, just because a criminal act created the conditions for the transaction.
Moreover, the civil court generally has an obligation to uphold the honest interests of every party that acted in good faith. The buyer of the house or the bank are honest, unsuspecting victims, just like the house's original owner. The law can't just say "Screw the buyer or bank!" and leave them out in the cold.
To an intelligent lay person, this all seems absurd, and rightly so. It demands constant vigilance of all homeowners, every day, every hour. A crook might do it during your two-week vacation, even--could it even happen while you're at work for the day? It seems much more reasonable to burden the buyer or bank--while it requires increased caution, it's limited to the period of their investigation. They can spot the fraud with homework: compare signatures, visit the neighbors, background check the seller/mortgage-holder, etc. When it's done, they can stop worrying about it.
It's like King Solomon and splitting the baby--the legal answer and the intuitively "right" answer are quite different. Lucky for the honest homeowners, it's very likely that as judges and lawyers become aware of these problems and reflect on them further, subsequent case law will start to get it right. Or the legislatures might change the law, but I wouldn't bet on that, first.
How thin is the air, up there where you're at, that you somehow believe that they wouldn't be allowed to present? Why is that "tough"
Since when does the Canadian government ask whether there is a "genuine use of [a] technology without breaking the law" before they pre-emptively restrict free speech? I'm pretty sure that they don't--go wikipedia it, yourself, and come back and tell me if I'm wrong, OK?
So where did you get this idea that somebody could stop their presentation/publishing?
* You may be confused by certain past cases (such as the RIAA/MPAA watermarking contest) wherein researchers are threatened with lawsuits by other private parties on contractual or copyright-related grounds. Zero application, here--these researchers aren't involved with any 2nd parties who have the legal standing and desire to bring such a tort.
* You may also be confused by the DMCA, or its counterparts in other countries, which criminalize the distribution of devices or methods that circumvent copyright protection mechanisms, like DVD's CSS encryption. Again, zero application, because this has nothing to do with copyright law.
* Is it possible that you were thinking of how governments will classify research that has national security implications, such as work on nuclear weapons or cryptography, muzzling the researchers with threats of criminal prosecution? Again, not an issue here--Faraday's law of induction isn't what you'd call a national secret.
So... Seriously: Am I missing something, here? Why DO you think these researchers would be stopped from presenting? And who do you think would do it, and how?
What you've described is a theoretical risk that basically never happens in the real world. I would guess that it's mostly because of the comparative risks of getting caught, versus the alternative means of getting data out the door.
First of all, racked servers are generally behind a lock door in a restricted area. Whether its fear of data theft, equipment theft, or just downtime due to a careless janior unplugging a server, most companies are smart enough to know that you don't want random juggalos with physical access to your racks. So the only people with access are management and IT.
Second, pulling drives out of running servers tends to get noticed, fast. Even if it's a fully mirrored RAID array, the monitoring software will probably notice and start complaining loudly to the on-call alert address, or wherever your alerts go. If it's not mirrored, well, would you mind stopping by the helpdesk and opening a trouble ticket on your way out?
So even if you had access to the servers, why would you physically steal the drives? It's highly visible so it'll be noticed quickly, and the pool of potential suspects is really small. I would posit that there is a strong correlation between those two factors and jail.
Here's a more likely alternative: Plug an innocuous USB thumb drive into the back of your workstation and copy the data from the server drive to your USB stick, over the network. Then walk out the door. If necessary, alter logs, etc. as needed to cover your tracks. The company might never find out about the theft, and even if they did, they'd have to pull you out of a larger pool of suspects. You'll probably get to keep your job.
(Bonus lulz: Alter the logs to frame your jackass boss, but only after spending a couple of weeks sucking up to HIS boss, so that you get promoted when he gets the v&.)
Patrick, did you read the second paragraph in the GGP? If not, go back and look at this quote:
"First, as an object approaches C the time dilation effect becomes such that from a frame of reference of the origin, the object never in fact can reach its destination. Would it not become in essence stuck in time?"
This is nonsensical in special relativity. In the frame of reference at the spacecraft's origin (Earth), the spacecraft will certainly reach its destination eventually as long as it has a positive velocity on that course. Time dilation's got nuthin to do widdit, there is no "time dilation effect" on our perception of how long it takes for the craft to reach its destination. If the craft moves with an average speed of 0.1c, it will take 10 years to reach a star that is 1 light-year from Earth. Period.
And WTF about the "stuck in time" part? Maybe he's thinking of how an observer perceives an object travelling toward the event horizon of a black hole. In that scenario, the object appears to slow down asymptotically as it approaches to the event horizon, never actually reaching it, at least to an outside observer.
You make a good point, in your post. In the GGP's case, though, I think he's on a completely different track.
From your post, you don't make it 100% clear, but I suspect your understanding of time dilation might not be 100% accurate.
Say the distance from Earth to another star is 1 light-year, and we manage to accelerate a probe to an average speed of 0.1*c (1/10th the speed of light). For the sake of our thought experiment, let's assume the probe comes back, too, for a total trip distance of 2 light-years.
On earth, 20 years will have passed--it's a simple, easy "distance = rate * time" kind of thing. No time dilation to consider.
If you placed a clock on the spaceship, though, you'd see some time dilation effects on the moving clock. It would have experienced less than 20 years' worth of time passing. So if your Earth-bound clock and your space clock were perfect, and you synced them up before the trip started, they would be out of sync when the ship got back.
Remember, in your own reference frame, you don't experience any time dilation. The fact that the ship is travelling fast doesn't make clocks on Earth run slower.
If this isn't clear, go read the Wikipedia article on time dilation, and read the part where it talks about muons decaying as they travel from the upper atmosphere to the surface of the Earth. That's the easiest example to understand, I think, as long as you get how radioactive decay operates.
If you'd bothered to even finish reading the summary (let alone the article), you would have noticed the key word: SHARED. Nobody's talking about hosting this all on one physical computer any more than Gmail is hosted on one physical computer. Both setups are distributed clusters of smaller computers.
At which point you start to see were IBM's idea actually make sense--they are talking about building a worldwide, distributed, networked collection of cooperating computers... HEY, that sounds an awful lot like the Internet!!
(I swear, the comment quality on Slashdot gets more and more like YouTube every day.)
Not to needle you, but I'm bored and this is Slashdot, so I'm gonna needle you.
The term "mixture" implies that there is some kind of possible theory of intelligence that contains elements of both of the original theories expounded by the article: That any given person's initial level of intelligence IS static and CANNOT be improved through effort, and that any given person's initial level of intelligence IS NOT static and CAN be improved through effort.
How do you think it's possible to mix those two theories into one? They only differ in a single element, which is a very straightforward binary choice: Either the human brain can get smarter through effort or it can't. I would posit to you, logically, that it is not possible to mix those two theories together and have a logically consistent statement.
Maybe you can try to explain your mixed theory, again. Like I said, I'm pretty sure you either misunderstood the article or didn't read it.
Your first two sentences directly contradict each other. The first clause in your second sentence is logically meaningless--I'm betting that you either totally misunderstood the article's point, or that you didn't read it.
Here, in the second sentance, you appear to be implying that intelligence starts at some "natural" level (whatever that's supposed to mean) which is in turn malleable and improved through effort and challenge:
To me, all people have a level of natural intelligence, that can be both improved and extended through hard work and challenging the brain.
But that's ONE of the two competing theories of the mind that the article talks about. The other theory is that intelligence starts at some natural level, and but that it cannot be improved by effort and hard work. Which is why it's so confusing that your first sentence:
I tend to think both theories of intelligence are true.
claims that you agree with both theories. But, obviously, you do not--your first sentence is disproven by your second sentence.
No, your point is that you've devised a tautology and are now bragging that nobody can disprove you. That's such an accomplishment--you must be a truly great mind, and to think you're here on Slashdot talking to me!
Market economies, by definition, MUST contain regulation. There are anarcho-capitalists who would lead you to believe otherwise, but their basic thesis (that reputation effects, alone, can provide enough regulation to allow markets to be efficient) has been generally disproven in most markets. Regulation could be limited to just basic criminal law and contract enforcement, or it could include more sophisticated regulatory bodies like the SEC, FCC, or FTC.
Nobody who uses the term capitalism actually intends the meaning that you're assigning, of a "real free market". That's why I'm accusing you of tautological reasoning: You've defined "capitalism" to mean some kind of fictional, theoretical construct that has no relation to either the real world or to the study of market economics. Since you control the definition, of course you can prove a point about it.
You should study some economics, learn at least a little about the basic terminology and theory, and then come back and we'll talk. I think my work is done, here, for now.
No, Iranians aren't allowed because U.S. law forbids commerce with residents of that country. Google doesn't want its employees going to jail or getting fined for breaking the law.
No, he wasn't right, because NONE OF THOSE NATIONS WERE ACTUAL MARKET CAPITALISM. They had some traits that are necessary components of markets, but they lacked a lot of other important traits that are also necessary. Economists studying the transitions of the former Eastern Bloc nations generally agree that some of these missing features were:
* A consistent and sensible rule of law to encourage good business practices, enforce contracts, and identify and punish fraud.
* A generally high level of "market experience" on the part of market participants, OR a flexible, honest, transparent system of government regulations to rein in business practices and level the playing field between naive consumers and wily businessmen.
The situation in the Eastern Bloc was pretty anarchic, and the general public had basically no market experience. The only people who did have market experience were those who'd participated in the black market extensively under the old system, and a lot of those guys weren't exactly honest citizens. Contrast that with nations that have largely successful markets, like the US or the UK: the consumers are pretty well educated, and in markets wherein the public is considered somewhat naive (like investing) there are a LOT of regulations that seek to level the playing field. Fraud is generally punished, too.
But, more generally, nobody can point you to a country based on a "true free market", because there is no such country. Some nations call themselves capitalist and some call themselves socialist and some call themselves communists. But very few nations are pure, textbook examples of these types of economic systems, or even close. Here's an illustrative example:
In the 1970s, Richard Nixon (the President of the USA at the time) led the US Congress to pass laws implementing pretty strict wage and price freezes on large sectors of the economy, in a misguided attempt to control inflation (it didn't work). In the 1980s and beyond, the US Federal budget ballooned to a huge fraction of the total economy, as expressed in the GDP (largely due to military spending). In fact, since WWII, the economy of the United States has been controlled to a considerable extent by government spending decisions, regulations, and other direct and indirect controls.
Contract that picture with the Soviet Union between 1950 and 1990. During the 1970s and 1980s, Western economists estimated the sizes of various communist countries' economies based on those countries' own published production data. These same economists considered estimating the sizes of the black market components, as well, but lacked any real data on the subject, so there were no solid conclusions. Most people figured that the Soviet black market was no more than 1-5% of the total economy.
After the wall came down, those same economists started getting reports of factories in East Germany, Poland, and Russia that had been idle and derelict for thirty years, but which figured prominently in industrial-production reports. Basically, the official figures were overstated by as much as 30-50%, between 1965 and 1990. In contrast, new data and analysis about the black markets also emerged, and some pretty solid numbers emerged, which put the size of the black market at more than 10x the earlier estimates. In short, the black market (capitalist) made up anywhere from 1/5 to 1/2 of the entire Soviet economy. Crazy, eh?
So despite the rhetoric, the US and the Soviet Union had a lot more in common, economically, than is generally supposed. The big differences lay in the methods of government: The US had, comparatively, a lot more political participation, civil liberty protection, and economic mobility than did the Soviet Union. Even during the 1950s, when anticommunist fervor led to some pretty nasty chilling effects on political freedom, things were still a lot more open than under Stalin during the same period.
It's important not to make the mistake of confusing the organization of the economy (communist, socialist, capitalist) with the organization of the government (authoritarian, democratic) or the methods of the government.
Um... Branko Horvat was a socialist politician. He was anti-communist, and then when the Iron Curtain came down, his first (and only) experience with capitalism was watching the looting, cronyism, and mafia tactics that accompanying six years of perestroika followed by the dissolution of the Soviet Union.
The experience in most of the former Eastern Bloc countries was similar: Basically, the economy got raped by opportunists and swindlers, and neither well-meaning but naive politicians nor careful by naive publics stood a chance. Eventually, in Russia, the looters and swindlers hooked their money back up with the elements who'd run the old authoritarian government (ex-KGB bureaucrats) and the bulk of the economy (black-marketeer organized crime). The new crew has spent the last ten years working to rebuild the old authoritarian government, calling it capitalism but using the power of the state to rob anybody (e.g., Gazprom) who wouldn't take a seat at their table.
Croatia was arguably subject to even more economic devastation, less due to the looting and more to the war. Since the end of the war, politics and business have been less obviously corrupt than Russia, but not without their share of problems. Since Horvat died in 2003, I doubt his personal perspective had much of a view beyond the war years--but I may be wrong about that.
So as I'm sure you can see, the logical problem with quoting Horvat on capitalism is that he had no direct experience with actual capitalism, in his lifetime. Sure, they called it capitalism, in the 1990s, but the region was either at war or in the midst of a massive transitional state at the time. For an analogous situation, why not ask the ghost of Trotsky what he thinks of the Stalin years? Is it fair to call that communism, or was it more of a corrupt dictatorship run by a madman in the grip of fierce paranoia? Or how about Germany under the Third Reich--they called it "National Socialism", but really it was another dictatorship.
The point is, any reasonable observer would conclude that the economies of some modern nations that call themselves market capialists (like the United States) are vastly different in practice than those of other nations that use similar names (Russia). Horvat didn't make that distinction--he was searching for reasons why his country had failed, and he ended up blaming the market when the real culprit was naivety and a populace robbed of their independent thinking by decades of authoritarian rule.
Or, you could use per-block encryption to allow random seeking on file access operations, with block granularity. You might think that block granularity isn't true random access, but it's what hard drives do (you have to read/write from the drive in terms of 512-byte sectors, not individual bytes).
This method has a side effect of reducing the amount of side-channel information that a server-side spook-installed tap can gather. He'll see your access patterns, in terms of whether you're reading and writing small or large chunks, whether your access is largely contigous or not, etc. But he doesn't know which block belongs to which file, so this information is of limited usefullness to him, aggravated by the amount of fragmentation in the "partition".
Actually, you could implement a model of this right now: export an iSCSI volume on the "server" side, and mount it on a Linux client that supports some kind of block-level encryption. This doesn't guarantee the data integrity against third-party tampering, so you'd probably want to add IPSEC IKE to the connection, too, which is minimally performance-impacting.
I'd like to see some benchmarks of these alien-technology SSDs you're using. I have worked with RAM-based SSDs that could push 200 MB/sec, sustained, of data across a PCI bus, and with zero device latency--the total latency of operations against the SSD was the round-trip time on the PCI bus. That's significantly faster than any single hard drive on the market.
I just spent a few minutes checking prices, specs, and reviews for solid-state disks on NewEgg, which has a pretty good selection. You can get SSDs for $30/GB that have sustained transfer as fast as conventional disks. (For comparison, actual disks are ~ $0.30/GB, and DDR2 RAM is ~ 20/GB.)
These prices point out that you're distorting the truth, too: Some flash SSDs do exist that are as fast as a moderately speedy hard disk, but these devices are insanely expensive, far outside the range of most buyers. I mean, seriously, your "fast" SSDs are barely pusing 45 MB/sec sustained writes (compared to upwards of 85 MB/sec for 7200 RPM perpendicular-recording hard drives), and the 64 GB version costs $4,000.00. Did you notice, also, that DDR2 RAM costs only $20/GB, which means that adding more physical RAM is less expensive than adding swap on a flash disk! (The slow, cheap SSDs, FTR, are basically CF cards with a laptop IDE interface, and are much slower than conventional drives, so they don't even enter into this discussion--good luck trying to swap on one of those!)
So which adds more to performance, increasing physical RAM by 1 GB or increasing swap by 1 GB? Duh. Given also that RAM is cheaper than swap on SSD, per GB, shouldn't I buy the faster one? And remember, physical RAM is a LOT faster than swap on SSD--like, 1000x less latency and more throughput. Which returns us to my original point, that you so handily ignored: You'll get a much more bang for your buck out of buying more RAM than you will adding swap.
There might be a performance gain to adding a fast SSD for swap, but only if we're dealing with a situation where you can't slot more RAM in your system. The RAM limit is about 4GB on modern laptops, maybe more on desktops. In that case, since you *can't* grow your virtual memory pool by adding physical RAM, swap would be the only option. But you make a lot of exaggerated claims about how much impact adding virtual memory beyond 4GB will have.
Unless your total working set (on-disk files, on-disk programs/libraries, running programs etc.) consists of more than 4GB, your proposed 32GB swap partition will have exactly zero impact on system performance. That's right, NONE. And looking at my Windows workstation, right now, shows that even including:
* all the ordinary garbage you'd expect on Windows XP Pro
* ~500MB of source code and binaries in my IDE
* Firefox with 10 tabs
* Thunderbird with a Gmail IMAP session containing 10,000+ messages
* 2 moderately sized OpenOffice spreadsheets
* 4 OpenOffice documents at 15+ pages, each
* Eclipse with the aforementioned 500MB collection of code I'm still only using less than 3 GB, total, out of 4 GB of physical memory. And remember, the disk cache is included in that 3 GB. That's right--with all that bloated, memory-hogging software and 500MB of on-disk files in pretty heavy usage, I'm not swapping at all. My system is responsive, and I can jump around from Eclipse to Firefox to Thunderbird to OpenOffice without any latency or thrashing, at all. I just ain't doing enough to using more memory, no matter how hard I try.
The user experience doesn't get better than that, performance-wise. Sure, I could add the swap, and it would slowly accumulate more data in the page cache as I work, but I'm never accessing most of it so it doesn't help my performance. You would see allocated swap containing data that hadn't been used in hours, which doesn't actually improve my user experience at all because I never need to retrieve all that extra accumulated data--it's leftover from some DLL or INI file accessed during system startup.
So explain this to me, again, one more time: Why exactly should I spend $2,000 on a new SSD in 2008?
If it's a flash-based device and the access throughput rates are greater than the values I quoted above (20 MB/sec writes, 12 MB/sec reads), the device must be implementing some kind of RAID-like striping. So that could explain the benchmarks to which you linked. I use a similar strategy building CF-based server root drives: take 3-4 CF disks of the desired size attached to SATA converters, and then hook those up to a 4-port 3ware SATA RAID controller in a RAID1 (mirror) of RAID5 configuration. The 3ware has a management tool for periodically comparing mirror components to ensure that no silent data corruption has crept into the mix.
The idea, here, is that the root/boot drive of most servers has basically read-only access patterns (if you offload logging and app-specific stuff like databases to dedicated storage, via the network), and it's mostly random access. The performance of RAID1 or RAID5 array for reads and for random access is fantastic. Having the data integrity check and a hardware controller ensures that if you do start to experience flash cell failures related to the write cycle issue, you will notice them and the RAID controller will compensate.
It's a pretty cool configuration, overall--assuming, of course, that you have iSCSI or local RAID arrays to which to offload non-root/boot data.
1) Most of the SSD devices coming on the market right now are *not* just RAM-same-as-your-main-memory. We are talking about two very, VERY different technologies, here, with vastly different performance characteristics. I can't blame you for being confused, though--SSD is a bullshit marketing term, anyway, and as such it's intended to confuse the consumer.
A traditional SSD is a box containing sticks of SDRAM-like-your-main-memory (or EDO in older incarnations, or DDR in newer ones). These RAM chips are on sticks identical to what you put in your mainboard slots, and they are just as fast. Often, these are hooked up directly to the PCI bus or a SCSI bus, instead of to an IDE bus, because they actually have enough device-side I/O bandwidth to swamp your IDE bus and then some. However, this kind of SSD doesn't retain data with the power turned off for more than a few seconds, because the 'D' in SDRAM stands for 'dynamic'--it needs constant power refreshes to maintain state. Some even have built-in notebook hard drives and automatically copy data to and from the notebook drive when they lose or regain power.
A new-style SSD is box containing sticks of FLASH NAND memory, like in your digital camera memory cards or most low-capacity MP3 players. In fact, you can make your own with any old CF-format flash card and a CF-to-IDE or CF-to-SATA adapter card (google for them--they're about $10-$20 each)--this is because the CF and SD standards incorporate the ATA command set used by IDE and SATA disks, so the adapter really only needs to modify the bus characteristics. Some purpose-built SSDs claim to offer advanced wear-leveling techniques, and might also implement S.M.A.R.T. monitoring, which was left out of the CF and SD standards, but everything shows up as another hard drive, no matter. These guys are also as slow as flash memory, which currently maxes out around 20 MB/sec reading and 12 MB/sec writing, for the best reasonably-priced cards I've seen (SanDisk Ultra III), which is much slower than conventional hard drives (45 MB/sec reading/writing at the outer edge for conventional 7200 RPM disks, or up to 85 MB/sec for the newest perpendicular-recording disks with the proper firmware). But, they retain data for 5-10 years with the power off, which is about as good as you'd expect from hard disks (which tend to freeze up during years of inactivity).
So, in short:
(a) SSDs can be either RAM-backed or flash-backed. The former might possibly have a normal hard disk for power failure backup, too. (b) Depending on the type of SSD, they can either be faster than your PCI bus (RAM-backed) or pretty slow (flash-backed). (c) Each type has different weaknesses: flash-backed devices have limited write-cycle endurance issues, while RAM-backed devices have the power failure issue. (d) The prices are vastly different: flash-backed devices cost a few hundred dollars, or you can make your own for the cost of a CF card plus $20, while RAM-backed devices can get to be more than $10,000.
So don't go around saying that your main memory and an SSD are even approximately equivalent, because they're not.
2) If you have 2GB of RAM, why do you have *any* swap space configured? Swap is old-school, from back in the days when computers had 256MB or less of main memory. Tell you what, try this: On any heavily-used workstation or laptop you own, Linux OR Windows, that has 1GB of RAM or more, just disable swap entirely. Try it for a week or two--you won't notice a difference.
You see, people have been following these idiotic virtual memory suggestions about using "1.5x main memory size" from back in the old days, but they're meaningless, now. (Of course, anybody with a real hardware budget and some know-how would have told you that if you're swapping you need to buy more RAM.) And today, since the price of RAM is so low, we can all afford to just buy more RAM. Go to NewEgg and get a 1GB stick of DDR2 RAM for about $35, shipped.
I like how you put the word addiction in quotes, cleverly re-defining the word to mean something completely different from what the word *actually* means. And by that, I refer to the medical definition of addiction in the DSM-IV and the literature, which is used by psychologists and doctors (including psychiatrists) and pretty much the rest of the human race. Except you--you have a fondness for the methods of Lewis Carroll, perhaps?
I'll boil it down for you: Addiction is a psychological phenomenon wherein an addicted subject comes to focus on some external, directly reward-inducing activity to such a degree that it attenuates normal behavior, and the compulsion continues even in the face of negative consequences. In this usage, the term "normal" means the behaviors of the subject prior to the introduction of the external activity, but it also includes social and statistical norms, to some extent.
In other words, you're addicted when the particular activity or substance dominates your behavior, modifying the way that you live your life to a significant degree, and you don't stop the activity even when the bad consequences build up. An illustrative thought experiment is to take the addict and consider the opinion that his or her past self might have, looking forward in the future from the early days of use. If a guy who had maybe done coke once or twice had a vision of the future a la "A Christmas Carol", and saw himself five years later missing mortgage payments and losing his family because all he wanted to do is blow lines up his nose, how would he view the situation? Granted, this game doesn't model a lot of specific cases (such as adolescents), so take it with a grain of salt, but I think the point is pretty clear.
So, you see how the sex, air, and food examples don't really work, here. Consider that:
* I am generally receptive to sex, and sexual thoughts often enter my (male) mind.
* The act of sex causes me great pleasure when I engage in it.
* I would express disappointment at being denied sex, if I had some expectation of it at a particular time.
But also consider:
* I don't compulsively pursue sexual activity to the exclusion of working, socializing or engaging in other activities.
* My sexual activities don't cause large negative consequences to me, like my GF leaving me because I can't stop having casual sex with other women, or the cops picking me up for soliciting prostitutes.
* I'm pretty confident that my peers who share my general values regarding sexual activity (i.e., not hard-core Christians) would be OK trading patterns of sexual activity with me. Other people (caveat as given) don't generally look at my sexual behavior and go "Eew, that guy needs help."
Now, there ARE people who have sexual addictions, that have exactly that last set of problems I just mentioned. These aren't just people who like to have sex a lot, or who have high sex drives--these are people who are constantly trolling bars or cruising for hookers several nights a week, who lose their jobs after being warned about looking at porn at work and keep doing it. These are people who are ashamed of their actions, even in the company of generally sexually-liberated folks, and who often want to change their behavior but don't see a way out of it.
It's the same with booze, coke, heroin, gambling, cigarettes. Virtually everybody in college (in the US, anyway) engages in binge drinking, where you get blackout, puking drunk with your pals several nights a week for four straight years. Some of these people don't go to class, don't study, and fail out, while others finish up just fine (maybe not summa cum laude, but well enough) and graduate. When they leave college, some people grow up and start drinking more responsibly, while others keep doing it and end up sacrificing relationships, job performance, finances, etc. The bottom line is, some people who do it are addicted, and (usually) most people who do it aren't, a
Totally offtopic, but my freshman roommate and I were both military history buffs and slightly nutter. We had an imagination game called "Go back in time with a machine gun" that we used to play while drinking heavily. The challenge was to construct the most compelling fantasy of altering historical events merely by traveling back in time for a few hours with an M-60 machine gun and a lot of ammo. The Punic wars (for instance) just wouldn't be the same.
Although the fact is that the game only really makes sense up through roughly the Napoleonic Wars, because the impact of a single guy with a machine gun can't be that large after that. You might have an impact because you know the course that a battle will take, but we eventually considered that to be cheating--you're leveraging the element of surprise and your knowledge of the future, rather than your machine gun and your moxie.
I eventually adapted the game for the debate team, since we shared a lot of long road trips in cramped rental cars and needed distractions. But since there were a lot of women involved, and women don't really care too much for military history, we came up with an alternative: Go back in time with any object currently in your dorm room.
One of the best answers I ever heard was to take your laptop (with full batteries, of course) and a portable scanner back to Alexandria and start copying scrolls in the library like crazy, a few days before it burned to the ground. Somehow, it seems a little more satisfying than letting the Carthaginians conquer Rome, or kicking the Normans back to France in 1066.
Slashdot Mirror
"It's interesting to read how you so clearly hate the idea of someone else having control, and at the same time you want that exact same control all to yourself above the devs. That's the most common problem with the middle-management level of development, they're in charge and they're going to have it their way, all other factors be damned."
I AM a dev, you idiot. Back when I was fresh and new and guileless, like you, I thought a lot like you. Everything I knew about development had come from two sources, CS assignments and Slashdot--I didn't even know how little I knew.
Most of that got that beat out of me, eventually--not by management, but by other engineers who just didn't have the patience for my snotty, immature attitude.
"What the development team as a whole needs is balance. I'm not saying for either side to hold things hostage, I'm just saying don't let the PHB's decide when it's "ready" based solely on a trade show date."
And now you're just redefining your side of the argument because I burned it out from under you, in my previous comment. Changing your tune, now, cannot erase your previous comments. YHL, HAND.
"...athough you gave a reasonable example of how there is only a finite range of shoelace lengths that are practical..."
NO. That's NOT what I did.
I demonstrated that:
* While there is a THEORETICALLY INFINITE RANGE of possible theoretical shoelace lengths;
* Nonetheless, the probability distribution of actual shoelace lengths assigns a non-infinitessimal (and rather large) probability to a finite (and rather small) range of lengths.
Do you remember WHY that's important? It's because in your previous post, YOU CLAIMED that:
"...in an infinite field of possibilities, everything is infinitesimally rare."
Which is just utter horseshit. You denied the existence of uneven probability distributions--seriously, what the fuck is wrong with you?
You are now, officially, squirming like a fish on a hook. You have been owned at every comment, by pretty much everybody, probably because your math training is well below average for Slashdot.
You can keep trying to sneak your way out of this argument, so you can maintain your fragile sense of self-worth, but nothing will change the fact that you got your ass royally whooped.
Go back to school, kiddo, and pay more attention in class.
So... You're a perfectionist who will hold a project hostage until YOU decide, on your terms, that it's not "crap"? And anybody who disagrees with your assessment of costs versus benefits is a "pinhead"?
Remind me to hire you, really. I love programmers who sneer at business concerns, refuse to compromise with their teams, and are so inexperienced that they believe that any business can get done that way.
Man, if I had a whole team of guys like you, I'd be able to... Hmm. People like you actually detract from the productivity of a team. With enough of your clones running around bragging about how big their dicks are, business would grind to a halt. Hell, with a whole company of you, we could probably put global software productivity into reverse, it would be so bad.
"...if I had a really large project..."
I think this says it, right here, folks. If you ever you DO get on a large project, give me a call. Maybe you'll have learned enough to hire you.
Who said we were talking about MS and Windows? You just brought that up, right now. I don't think it proves anything, one way or another, that one company has a crappy process.
Honestly, it seems like you just tried to "move the goalposts", redefining the terms of an argument you were losing so you can feel like you're winning.
That's lame, and I'm calling you out on it.
You incorrectly assume that "an infinite number of different programs" and "all possible programs" are the same set. They are not.
Turing's proof shows that no algorithm can solve the Halting Problem for *all possible* programs. But there ARE proven algorithms that solve the the Halting Problem for certain classes of programs, that is, subsets of "all possible" programs.
Many of those subsets (all the interesting ones, really) containg an infinite number of possible programs. Not *all* possible programs, mind you--just an infinitely-large subset of another distinct, infinitely-large superset.
(Please read the Wikipedia page on infinity, and grok the part about different sizes of infinity, before we continue. Don't worry, I've got time.
Got it? Good.)
Also, you wrote "...in an infinite field of possibilities, everything is infinitesimally rare." This statement is mathematically absurd--obviously, you've never taken probability, so you don't know how to talk about it in a useful way. Let me demonstrate:
There are an infinite number of possible lengths--the entire real number space, right? But not all shoelace lengths are all equally probable in that space. I'd guess that more than 99% of all shoelaces are shorter than 1 meter--my local drug store doesn't stock anything longer than about 75 cm.
Now, does that mean that my drugstore's policy "can't work"? Of course not--it works fine, they make a profit. If a clown with giant shoes needs mile-long shoelaces, nobody cares. It's perfectly useful to only solve 99% of the possible cases, right?
Back to programs: IF you define the probability distribution of human programmer output as totally random, you'd have a point. But humans don't bang out random strings of gibberish machine instructions like monkeys at typewriters. We tend to follow certain well-worn paths, and some of these subsets have proven Halting Problem-solvable. Tools that implement those insights aren't "general algorithms", but they're still useful.
Go read up on tools like Electric Fence, and other automated debugging aides. This is not a new concept, and you should really learn more about the subjects (math, software). It's embarrassing, for you.
How large of a programming team do you work with? And how big are the projects to which you contribute code? And what kind of development model do you use (waterfall, Agile, ad-hoc, etc.)?
Shipping a large project with 1,000 bugs might be a perfectly valid decision. Are any of those 1,000 bugs deal-breakers for your install base? If so, how many clients does it affect? Are these "real bugs", or just incomplete/unpolished functions, or documentation issues, or output typos, or what?
And what kind of software is this? Are you building a time & expense web application, or a filesystem driver? In the former case, most bugs will be interface glitches--ugly, annoying, and harmless. In the latter case, even one bug could easily cause silent data corruption.
Remeber what Linus Torvalds said: Release early, release often. Don't wait til all your bugs are fixed before shipping your software, or you'll lose your "market" window. If it's good enough, the early-adopters will understand, and might even contribute bug reports or patches that will speed you up.
Has anybody every told you "'Perfect' is the enemy of 'good enough'."? Perhaps after listening to you explain why your project is behind schedule, then sighing and face-palming?
The halting problem says that there cannot be a GENERAL ALGORITHM that works in all cases, for any of the infinity of possible programs that can exist.
That proves ZERO about, say, whether I can write an algorithm that covers 99% of the common cases. The lack of a general solution doesn't imply that it can't be done often enough, in practice.
Kind of OT, but...
The simplest scheme in real estate fraud is to rent a house, fill out some fake land title transfer paperwork, and then impersonate the lawful owner and either (A) sell the house to an unsuspecting, honest buyer, or (B) take out a bank mortgage on the house. Either way, you walk off with cash-in-hand.
Read and learn: http://www.abanet.org/genpractice/newsletter/lawtrends/0709/realestate/realestatefraud101.html . Google for many more news stories.
The really nasty part, here, is that the original homeowner is, in legal terms, cold fucked. In the case of an outright sale, the unsuspecting buyer actually has a stronger title claim to the property than the original owner--fraud notwithstanding. If the crook mortgaged the house, that bank has a possible (though weaker) legal claim on seizing the collateral (house) if nobody makes the payments.
The fundamental legal problem: In the US and Canada, the criminal offense of fraud is a totally separate matter from the civil property issue (the title claims, or the mortgage). A subsequent civil transaction isn't automatically voided, just because a criminal act created the conditions for the transaction.
Moreover, the civil court generally has an obligation to uphold the honest interests of every party that acted in good faith. The buyer of the house or the bank are honest, unsuspecting victims, just like the house's original owner. The law can't just say "Screw the buyer or bank!" and leave them out in the cold.
To an intelligent lay person, this all seems absurd, and rightly so. It demands constant vigilance of all homeowners, every day, every hour. A crook might do it during your two-week vacation, even--could it even happen while you're at work for the day? It seems much more reasonable to burden the buyer or bank--while it requires increased caution, it's limited to the period of their investigation. They can spot the fraud with homework: compare signatures, visit the neighbors, background check the seller/mortgage-holder, etc. When it's done, they can stop worrying about it.
It's like King Solomon and splitting the baby--the legal answer and the intuitively "right" answer are quite different. Lucky for the honest homeowners, it's very likely that as judges and lawyers become aware of these problems and reflect on them further, subsequent case law will start to get it right. Or the legislatures might change the law, but I wouldn't bet on that, first.
How thin is the air, up there where you're at, that you somehow believe that they wouldn't be allowed to present? Why is that "tough"
Since when does the Canadian government ask whether there is a "genuine use of [a] technology without breaking the law" before they pre-emptively restrict free speech? I'm pretty sure that they don't--go wikipedia it, yourself, and come back and tell me if I'm wrong, OK?
So where did you get this idea that somebody could stop their presentation/publishing?
* You may be confused by certain past cases (such as the RIAA/MPAA watermarking contest) wherein researchers are threatened with lawsuits by other private parties on contractual or copyright-related grounds. Zero application, here--these researchers aren't involved with any 2nd parties who have the legal standing and desire to bring such a tort.
* You may also be confused by the DMCA, or its counterparts in other countries, which criminalize the distribution of devices or methods that circumvent copyright protection mechanisms, like DVD's CSS encryption. Again, zero application, because this has nothing to do with copyright law.
* Is it possible that you were thinking of how governments will classify research that has national security implications, such as work on nuclear weapons or cryptography, muzzling the researchers with threats of criminal prosecution? Again, not an issue here--Faraday's law of induction isn't what you'd call a national secret.
So... Seriously: Am I missing something, here? Why DO you think these researchers would be stopped from presenting? And who do you think would do it, and how?
What you've described is a theoretical risk that basically never happens in the real world. I would guess that it's mostly because of the comparative risks of getting caught, versus the alternative means of getting data out the door.
First of all, racked servers are generally behind a lock door in a restricted area. Whether its fear of data theft, equipment theft, or just downtime due to a careless janior unplugging a server, most companies are smart enough to know that you don't want random juggalos with physical access to your racks. So the only people with access are management and IT.
Second, pulling drives out of running servers tends to get noticed, fast. Even if it's a fully mirrored RAID array, the monitoring software will probably notice and start complaining loudly to the on-call alert address, or wherever your alerts go. If it's not mirrored, well, would you mind stopping by the helpdesk and opening a trouble ticket on your way out?
So even if you had access to the servers, why would you physically steal the drives? It's highly visible so it'll be noticed quickly, and the pool of potential suspects is really small. I would posit that there is a strong correlation between those two factors and jail.
Here's a more likely alternative: Plug an innocuous USB thumb drive into the back of your workstation and copy the data from the server drive to your USB stick, over the network. Then walk out the door. If necessary, alter logs, etc. as needed to cover your tracks. The company might never find out about the theft, and even if they did, they'd have to pull you out of a larger pool of suspects. You'll probably get to keep your job.
(Bonus lulz: Alter the logs to frame your jackass boss, but only after spending a couple of weeks sucking up to HIS boss, so that you get promoted when he gets the v&.)
Patrick, did you read the second paragraph in the GGP? If not, go back and look at this quote:
"First, as an object approaches C the time dilation effect becomes such that from a frame of reference of the origin, the object never in fact can reach its destination. Would it not become in essence stuck in time?"
This is nonsensical in special relativity. In the frame of reference at the spacecraft's origin (Earth), the spacecraft will certainly reach its destination eventually as long as it has a positive velocity on that course. Time dilation's got nuthin to do widdit, there is no "time dilation effect" on our perception of how long it takes for the craft to reach its destination. If the craft moves with an average speed of 0.1c, it will take 10 years to reach a star that is 1 light-year from Earth. Period.
And WTF about the "stuck in time" part? Maybe he's thinking of how an observer perceives an object travelling toward the event horizon of a black hole. In that scenario, the object appears to slow down asymptotically as it approaches to the event horizon, never actually reaching it, at least to an outside observer.
You make a good point, in your post. In the GGP's case, though, I think he's on a completely different track.
From your post, you don't make it 100% clear, but I suspect your understanding of time dilation might not be 100% accurate.
Say the distance from Earth to another star is 1 light-year, and we manage to accelerate a probe to an average speed of 0.1*c (1/10th the speed of light). For the sake of our thought experiment, let's assume the probe comes back, too, for a total trip distance of 2 light-years.
On earth, 20 years will have passed--it's a simple, easy "distance = rate * time" kind of thing. No time dilation to consider.
If you placed a clock on the spaceship, though, you'd see some time dilation effects on the moving clock. It would have experienced less than 20 years' worth of time passing. So if your Earth-bound clock and your space clock were perfect, and you synced them up before the trip started, they would be out of sync when the ship got back.
Remember, in your own reference frame, you don't experience any time dilation. The fact that the ship is travelling fast doesn't make clocks on Earth run slower.
If this isn't clear, go read the Wikipedia article on time dilation, and read the part where it talks about muons decaying as they travel from the upper atmosphere to the surface of the Earth. That's the easiest example to understand, I think, as long as you get how radioactive decay operates.
If you'd bothered to even finish reading the summary (let alone the article), you would have noticed the key word: SHARED. Nobody's talking about hosting this all on one physical computer any more than Gmail is hosted on one physical computer. Both setups are distributed clusters of smaller computers.
At which point you start to see were IBM's idea actually make sense--they are talking about building a worldwide, distributed, networked collection of cooperating computers... HEY, that sounds an awful lot like the Internet!!
(I swear, the comment quality on Slashdot gets more and more like YouTube every day.)
Not to needle you, but I'm bored and this is Slashdot, so I'm gonna needle you.
The term "mixture" implies that there is some kind of possible theory of intelligence that contains elements of both of the original theories expounded by the article: That any given person's initial level of intelligence IS static and CANNOT be improved through effort, and that any given person's initial level of intelligence IS NOT static and CAN be improved through effort.
How do you think it's possible to mix those two theories into one? They only differ in a single element, which is a very straightforward binary choice: Either the human brain can get smarter through effort or it can't. I would posit to you, logically, that it is not possible to mix those two theories together and have a logically consistent statement.
Maybe you can try to explain your mixed theory, again. Like I said, I'm pretty sure you either misunderstood the article or didn't read it.
Your first two sentences directly contradict each other. The first clause in your second sentence is logically meaningless--I'm betting that you either totally misunderstood the article's point, or that you didn't read it.
Here, in the second sentance, you appear to be implying that intelligence starts at some "natural" level (whatever that's supposed to mean) which is in turn malleable and improved through effort and challenge:
To me, all people have a level of natural intelligence, that can be both improved and extended through hard work and challenging the brain.
But that's ONE of the two competing theories of the mind that the article talks about. The other theory is that intelligence starts at some natural level, and but that it cannot be improved by effort and hard work. Which is why it's so confusing that your first sentence:
I tend to think both theories of intelligence are true.
claims that you agree with both theories. But, obviously, you do not--your first sentence is disproven by your second sentence.
No, your point is that you've devised a tautology and are now bragging that nobody can disprove you. That's such an accomplishment--you must be a truly great mind, and to think you're here on Slashdot talking to me!
Market economies, by definition, MUST contain regulation. There are anarcho-capitalists who would lead you to believe otherwise, but their basic thesis (that reputation effects, alone, can provide enough regulation to allow markets to be efficient) has been generally disproven in most markets. Regulation could be limited to just basic criminal law and contract enforcement, or it could include more sophisticated regulatory bodies like the SEC, FCC, or FTC.
Nobody who uses the term capitalism actually intends the meaning that you're assigning, of a "real free market". That's why I'm accusing you of tautological reasoning: You've defined "capitalism" to mean some kind of fictional, theoretical construct that has no relation to either the real world or to the study of market economics. Since you control the definition, of course you can prove a point about it.
You should study some economics, learn at least a little about the basic terminology and theory, and then come back and we'll talk. I think my work is done, here, for now.
No, Iranians aren't allowed because U.S. law forbids commerce with residents of that country. Google doesn't want its employees going to jail or getting fined for breaking the law.
No, he wasn't right, because NONE OF THOSE NATIONS WERE ACTUAL MARKET CAPITALISM. They had some traits that are necessary components of markets, but they lacked a lot of other important traits that are also necessary. Economists studying the transitions of the former Eastern Bloc nations generally agree that some of these missing features were:
* A consistent and sensible rule of law to encourage good business practices, enforce contracts, and identify and punish fraud.
* A generally high level of "market experience" on the part of market participants, OR a flexible, honest, transparent system of government regulations to rein in business practices and level the playing field between naive consumers and wily businessmen.
The situation in the Eastern Bloc was pretty anarchic, and the general public had basically no market experience. The only people who did have market experience were those who'd participated in the black market extensively under the old system, and a lot of those guys weren't exactly honest citizens. Contrast that with nations that have largely successful markets, like the US or the UK: the consumers are pretty well educated, and in markets wherein the public is considered somewhat naive (like investing) there are a LOT of regulations that seek to level the playing field. Fraud is generally punished, too.
But, more generally, nobody can point you to a country based on a "true free market", because there is no such country. Some nations call themselves capitalist and some call themselves socialist and some call themselves communists. But very few nations are pure, textbook examples of these types of economic systems, or even close. Here's an illustrative example:
In the 1970s, Richard Nixon (the President of the USA at the time) led the US Congress to pass laws implementing pretty strict wage and price freezes on large sectors of the economy, in a misguided attempt to control inflation (it didn't work). In the 1980s and beyond, the US Federal budget ballooned to a huge fraction of the total economy, as expressed in the GDP (largely due to military spending). In fact, since WWII, the economy of the United States has been controlled to a considerable extent by government spending decisions, regulations, and other direct and indirect controls.
Contract that picture with the Soviet Union between 1950 and 1990. During the 1970s and 1980s, Western economists estimated the sizes of various communist countries' economies based on those countries' own published production data. These same economists considered estimating the sizes of the black market components, as well, but lacked any real data on the subject, so there were no solid conclusions. Most people figured that the Soviet black market was no more than 1-5% of the total economy.
After the wall came down, those same economists started getting reports of factories in East Germany, Poland, and Russia that had been idle and derelict for thirty years, but which figured prominently in industrial-production reports. Basically, the official figures were overstated by as much as 30-50%, between 1965 and 1990. In contrast, new data and analysis about the black markets also emerged, and some pretty solid numbers emerged, which put the size of the black market at more than 10x the earlier estimates. In short, the black market (capitalist) made up anywhere from 1/5 to 1/2 of the entire Soviet economy. Crazy, eh?
So despite the rhetoric, the US and the Soviet Union had a lot more in common, economically, than is generally supposed. The big differences lay in the methods of government: The US had, comparatively, a lot more political participation, civil liberty protection, and economic mobility than did the Soviet Union. Even during the 1950s, when anticommunist fervor led to some pretty nasty chilling effects on political freedom, things were still a lot more open than under Stalin during the same period.
It's important not to make the mistake of confusing the organization of the economy (communist, socialist, capitalist) with the organization of the government (authoritarian, democratic) or the methods of the government.
Um... Branko Horvat was a socialist politician. He was anti-communist, and then when the Iron Curtain came down, his first (and only) experience with capitalism was watching the looting, cronyism, and mafia tactics that accompanying six years of perestroika followed by the dissolution of the Soviet Union.
The experience in most of the former Eastern Bloc countries was similar: Basically, the economy got raped by opportunists and swindlers, and neither well-meaning but naive politicians nor careful by naive publics stood a chance. Eventually, in Russia, the looters and swindlers hooked their money back up with the elements who'd run the old authoritarian government (ex-KGB bureaucrats) and the bulk of the economy (black-marketeer organized crime). The new crew has spent the last ten years working to rebuild the old authoritarian government, calling it capitalism but using the power of the state to rob anybody (e.g., Gazprom) who wouldn't take a seat at their table.
Croatia was arguably subject to even more economic devastation, less due to the looting and more to the war. Since the end of the war, politics and business have been less obviously corrupt than Russia, but not without their share of problems. Since Horvat died in 2003, I doubt his personal perspective had much of a view beyond the war years--but I may be wrong about that.
So as I'm sure you can see, the logical problem with quoting Horvat on capitalism is that he had no direct experience with actual capitalism, in his lifetime. Sure, they called it capitalism, in the 1990s, but the region was either at war or in the midst of a massive transitional state at the time. For an analogous situation, why not ask the ghost of Trotsky what he thinks of the Stalin years? Is it fair to call that communism, or was it more of a corrupt dictatorship run by a madman in the grip of fierce paranoia? Or how about Germany under the Third Reich--they called it "National Socialism", but really it was another dictatorship.
The point is, any reasonable observer would conclude that the economies of some modern nations that call themselves market capialists (like the United States) are vastly different in practice than those of other nations that use similar names (Russia). Horvat didn't make that distinction--he was searching for reasons why his country had failed, and he ended up blaming the market when the real culprit was naivety and a populace robbed of their independent thinking by decades of authoritarian rule.
Or, you could use per-block encryption to allow random seeking on file access operations, with block granularity. You might think that block granularity isn't true random access, but it's what hard drives do (you have to read/write from the drive in terms of 512-byte sectors, not individual bytes).
This method has a side effect of reducing the amount of side-channel information that a server-side spook-installed tap can gather. He'll see your access patterns, in terms of whether you're reading and writing small or large chunks, whether your access is largely contigous or not, etc. But he doesn't know which block belongs to which file, so this information is of limited usefullness to him, aggravated by the amount of fragmentation in the "partition".
Actually, you could implement a model of this right now: export an iSCSI volume on the "server" side, and mount it on a Linux client that supports some kind of block-level encryption. This doesn't guarantee the data integrity against third-party tampering, so you'd probably want to add IPSEC IKE to the connection, too, which is minimally performance-impacting.
I'd like to see some benchmarks of these alien-technology SSDs you're using. I have worked with RAM-based SSDs that could push 200 MB/sec, sustained, of data across a PCI bus, and with zero device latency--the total latency of operations against the SSD was the round-trip time on the PCI bus. That's significantly faster than any single hard drive on the market.
I just spent a few minutes checking prices, specs, and reviews for solid-state disks on NewEgg, which has a pretty good selection. You can get SSDs for $30/GB that have sustained transfer as fast as conventional disks. (For comparison, actual disks are ~ $0.30/GB, and DDR2 RAM is ~ 20/GB.)
These prices point out that you're distorting the truth, too: Some flash SSDs do exist that are as fast as a moderately speedy hard disk, but these devices are insanely expensive, far outside the range of most buyers. I mean, seriously, your "fast" SSDs are barely pusing 45 MB/sec sustained writes (compared to upwards of 85 MB/sec for 7200 RPM perpendicular-recording hard drives), and the 64 GB version costs $4,000.00. Did you notice, also, that DDR2 RAM costs only $20/GB, which means that adding more physical RAM is less expensive than adding swap on a flash disk! (The slow, cheap SSDs, FTR, are basically CF cards with a laptop IDE interface, and are much slower than conventional drives, so they don't even enter into this discussion--good luck trying to swap on one of those!)
So which adds more to performance, increasing physical RAM by 1 GB or increasing swap by 1 GB? Duh. Given also that RAM is cheaper than swap on SSD, per GB, shouldn't I buy the faster one? And remember, physical RAM is a LOT faster than swap on SSD--like, 1000x less latency and more throughput. Which returns us to my original point, that you so handily ignored: You'll get a much more bang for your buck out of buying more RAM than you will adding swap.
There might be a performance gain to adding a fast SSD for swap, but only if we're dealing with a situation where you can't slot more RAM in your system. The RAM limit is about 4GB on modern laptops, maybe more on desktops. In that case, since you *can't* grow your virtual memory pool by adding physical RAM, swap would be the only option. But you make a lot of exaggerated claims about how much impact adding virtual memory beyond 4GB will have.
Unless your total working set (on-disk files, on-disk programs/libraries, running programs etc.) consists of more than 4GB, your proposed 32GB swap partition will have exactly zero impact on system performance. That's right, NONE. And looking at my Windows workstation, right now, shows that even including:
* all the ordinary garbage you'd expect on Windows XP Pro
* ~500MB of source code and binaries in my IDE
* Firefox with 10 tabs
* Thunderbird with a Gmail IMAP session containing 10,000+ messages
* 2 moderately sized OpenOffice spreadsheets
* 4 OpenOffice documents at 15+ pages, each
* Eclipse with the aforementioned 500MB collection of code
I'm still only using less than 3 GB, total, out of 4 GB of physical memory. And remember, the disk cache is included in that 3 GB. That's right--with all that bloated, memory-hogging software and 500MB of on-disk files in pretty heavy usage, I'm not swapping at all. My system is responsive, and I can jump around from Eclipse to Firefox to Thunderbird to OpenOffice without any latency or thrashing, at all. I just ain't doing enough to using more memory, no matter how hard I try.
The user experience doesn't get better than that, performance-wise. Sure, I could add the swap, and it would slowly accumulate more data in the page cache as I work, but I'm never accessing most of it so it doesn't help my performance. You would see allocated swap containing data that hadn't been used in hours, which doesn't actually improve my user experience at all because I never need to retrieve all that extra accumulated data--it's leftover from some DLL or INI file accessed during system startup.
So explain this to me, again, one more time: Why exactly should I spend $2,000 on a new SSD in 2008?
If it's a flash-based device and the access throughput rates are greater than the values I quoted above (20 MB/sec writes, 12 MB/sec reads), the device must be implementing some kind of RAID-like striping. So that could explain the benchmarks to which you linked. I use a similar strategy building CF-based server root drives: take 3-4 CF disks of the desired size attached to SATA converters, and then hook those up to a 4-port 3ware SATA RAID controller in a RAID1 (mirror) of RAID5 configuration. The 3ware has a management tool for periodically comparing mirror components to ensure that no silent data corruption has crept into the mix.
The idea, here, is that the root/boot drive of most servers has basically read-only access patterns (if you offload logging and app-specific stuff like databases to dedicated storage, via the network), and it's mostly random access. The performance of RAID1 or RAID5 array for reads and for random access is fantastic. Having the data integrity check and a hardware controller ensures that if you do start to experience flash cell failures related to the write cycle issue, you will notice them and the RAID controller will compensate.
It's a pretty cool configuration, overall--assuming, of course, that you have iSCSI or local RAID arrays to which to offload non-root/boot data.
A couple of things.
1) Most of the SSD devices coming on the market right now are *not* just RAM-same-as-your-main-memory. We are talking about two very, VERY different technologies, here, with vastly different performance characteristics. I can't blame you for being confused, though--SSD is a bullshit marketing term, anyway, and as such it's intended to confuse the consumer.
A traditional SSD is a box containing sticks of SDRAM-like-your-main-memory (or EDO in older incarnations, or DDR in newer ones). These RAM chips are on sticks identical to what you put in your mainboard slots, and they are just as fast. Often, these are hooked up directly to the PCI bus or a SCSI bus, instead of to an IDE bus, because they actually have enough device-side I/O bandwidth to swamp your IDE bus and then some. However, this kind of SSD doesn't retain data with the power turned off for more than a few seconds, because the 'D' in SDRAM stands for 'dynamic'--it needs constant power refreshes to maintain state. Some even have built-in notebook hard drives and automatically copy data to and from the notebook drive when they lose or regain power.
A new-style SSD is box containing sticks of FLASH NAND memory, like in your digital camera memory cards or most low-capacity MP3 players. In fact, you can make your own with any old CF-format flash card and a CF-to-IDE or CF-to-SATA adapter card (google for them--they're about $10-$20 each)--this is because the CF and SD standards incorporate the ATA command set used by IDE and SATA disks, so the adapter really only needs to modify the bus characteristics. Some purpose-built SSDs claim to offer advanced wear-leveling techniques, and might also implement S.M.A.R.T. monitoring, which was left out of the CF and SD standards, but everything shows up as another hard drive, no matter. These guys are also as slow as flash memory, which currently maxes out around 20 MB/sec reading and 12 MB/sec writing, for the best reasonably-priced cards I've seen (SanDisk Ultra III), which is much slower than conventional hard drives (45 MB/sec reading/writing at the outer edge for conventional 7200 RPM disks, or up to 85 MB/sec for the newest perpendicular-recording disks with the proper firmware). But, they retain data for 5-10 years with the power off, which is about as good as you'd expect from hard disks (which tend to freeze up during years of inactivity).
So, in short:
(a) SSDs can be either RAM-backed or flash-backed. The former might possibly have a normal hard disk for power failure backup, too.
(b) Depending on the type of SSD, they can either be faster than your PCI bus (RAM-backed) or pretty slow (flash-backed).
(c) Each type has different weaknesses: flash-backed devices have limited write-cycle endurance issues, while RAM-backed devices have the power failure issue.
(d) The prices are vastly different: flash-backed devices cost a few hundred dollars, or you can make your own for the cost of a CF card plus $20, while RAM-backed devices can get to be more than $10,000.
So don't go around saying that your main memory and an SSD are even approximately equivalent, because they're not.
2) If you have 2GB of RAM, why do you have *any* swap space configured? Swap is old-school, from back in the days when computers had 256MB or less of main memory. Tell you what, try this: On any heavily-used workstation or laptop you own, Linux OR Windows, that has 1GB of RAM or more, just disable swap entirely. Try it for a week or two--you won't notice a difference.
You see, people have been following these idiotic virtual memory suggestions about using "1.5x main memory size" from back in the old days, but they're meaningless, now. (Of course, anybody with a real hardware budget and some know-how would have told you that if you're swapping you need to buy more RAM.) And today, since the price of RAM is so low, we can all afford to just buy more RAM. Go to NewEgg and get a 1GB stick of DDR2 RAM for about $35, shipped.
Remember, givin
I like how you put the word addiction in quotes, cleverly re-defining the word to mean something completely different from what the word *actually* means. And by that, I refer to the medical definition of addiction in the DSM-IV and the literature, which is used by psychologists and doctors (including psychiatrists) and pretty much the rest of the human race. Except you--you have a fondness for the methods of Lewis Carroll, perhaps?
I'll boil it down for you: Addiction is a psychological phenomenon wherein an addicted subject comes to focus on some external, directly reward-inducing activity to such a degree that it attenuates normal behavior, and the compulsion continues even in the face of negative consequences. In this usage, the term "normal" means the behaviors of the subject prior to the introduction of the external activity, but it also includes social and statistical norms, to some extent.
In other words, you're addicted when the particular activity or substance dominates your behavior, modifying the way that you live your life to a significant degree, and you don't stop the activity even when the bad consequences build up. An illustrative thought experiment is to take the addict and consider the opinion that his or her past self might have, looking forward in the future from the early days of use. If a guy who had maybe done coke once or twice had a vision of the future a la "A Christmas Carol", and saw himself five years later missing mortgage payments and losing his family because all he wanted to do is blow lines up his nose, how would he view the situation? Granted, this game doesn't model a lot of specific cases (such as adolescents), so take it with a grain of salt, but I think the point is pretty clear.
So, you see how the sex, air, and food examples don't really work, here. Consider that:
* I am generally receptive to sex, and sexual thoughts often enter my (male) mind.
* The act of sex causes me great pleasure when I engage in it.
* I would express disappointment at being denied sex, if I had some expectation of it at a particular time.
But also consider:
* I don't compulsively pursue sexual activity to the exclusion of working, socializing or engaging in other activities.
* My sexual activities don't cause large negative consequences to me, like my GF leaving me because I can't stop having casual sex with other women, or the cops picking me up for soliciting prostitutes.
* I'm pretty confident that my peers who share my general values regarding sexual activity (i.e., not hard-core Christians) would be OK trading patterns of sexual activity with me. Other people (caveat as given) don't generally look at my sexual behavior and go "Eew, that guy needs help."
Now, there ARE people who have sexual addictions, that have exactly that last set of problems I just mentioned. These aren't just people who like to have sex a lot, or who have high sex drives--these are people who are constantly trolling bars or cruising for hookers several nights a week, who lose their jobs after being warned about looking at porn at work and keep doing it. These are people who are ashamed of their actions, even in the company of generally sexually-liberated folks, and who often want to change their behavior but don't see a way out of it.
It's the same with booze, coke, heroin, gambling, cigarettes. Virtually everybody in college (in the US, anyway) engages in binge drinking, where you get blackout, puking drunk with your pals several nights a week for four straight years. Some of these people don't go to class, don't study, and fail out, while others finish up just fine (maybe not summa cum laude, but well enough) and graduate. When they leave college, some people grow up and start drinking more responsibly, while others keep doing it and end up sacrificing relationships, job performance, finances, etc. The bottom line is, some people who do it are addicted, and (usually) most people who do it aren't, a
Totally offtopic, but my freshman roommate and I were both military history buffs and slightly nutter. We had an imagination game called "Go back in time with a machine gun" that we used to play while drinking heavily. The challenge was to construct the most compelling fantasy of altering historical events merely by traveling back in time for a few hours with an M-60 machine gun and a lot of ammo. The Punic wars (for instance) just wouldn't be the same.
Although the fact is that the game only really makes sense up through roughly the Napoleonic Wars, because the impact of a single guy with a machine gun can't be that large after that. You might have an impact because you know the course that a battle will take, but we eventually considered that to be cheating--you're leveraging the element of surprise and your knowledge of the future, rather than your machine gun and your moxie.
I eventually adapted the game for the debate team, since we shared a lot of long road trips in cramped rental cars and needed distractions. But since there were a lot of women involved, and women don't really care too much for military history, we came up with an alternative: Go back in time with any object currently in your dorm room.
One of the best answers I ever heard was to take your laptop (with full batteries, of course) and a portable scanner back to Alexandria and start copying scrolls in the library like crazy, a few days before it burned to the ground. Somehow, it seems a little more satisfying than letting the Carthaginians conquer Rome, or kicking the Normans back to France in 1066.
This is why I quit playing D&D--too inflexible.