I agree with you 100% that Wikipedia has peaked. The quality of most articles is dropping over time,
because anybody halfway sane doesn't want to pore autistically over a watchlist of cherished articles to make sure they don't succumb to entropy.
On the other hand, that doesn't mean that every dispute on WP is pointless, or that either side could be right on every issue. One bogus argument that's always posed by people who don't want their articles deleted is that it's not a paper encyclopedia, so there's no reason to keep the whole thing under a certain page count. Well, suppose Fred creates an article on his high school band, Fredsband, which only actually consisted of himself and his golden retriever. Every single time a user searches for "golden retriever," one of the hits is going to be the article on Fredsband. Also, when you have an article that's non-notable, it tends not to be linked to any other articles, and you get these little disjoint subsets of WP that are unhealthy. They can become havens for crackpots, or honeypots for spam links.
Here's a thought, use per machine accounts instead of per user accounts and click through the EULAs BEFORE imaging the drive.
That would break lots of other things we want (e.g., letting students read and write in their own networked folders), and it would be totally backwards to make decisions like that based on one piece of software.
I hope Sun will take a hint from this, and stop trying to impose the LGPL as a click-through EULA in the Windows versions of OpenOffice. I teach a physics lab course where I'm trying to encourage students to use OpenOffice instead of Office. (One really practical reason is that they'll make a graph using Excel at school, email it to themselves, then try to open it at home using Excel and find out they can't, because they have an older version of Excel at home.) The really annoying thing is that when you install OOo, it forces each user to click through the EULA the first time on that machine. This is lame, because:
The LGPL isn't a EULA, and doesn't require agreement from the end-user. You only have to agree to it if you want to redistribute it the sofwtare.
It's a hassle for my students. We have 7 Windows machines in the room, and I have 75 students in my lab classes. That means students potentially have to click through the license 75*7=525 times per semester.
But wait, there's more! These machines will soon be set up so that their hard disks are restored from a standard image every night at midnight. That guarantees that every student will have to click through the license every single time they start the software. That means if every student uses OOo once a week in lab, we potentially have the EULA being clicked through 1200 times per semester. Ugh!
Your academic papers don't have such a licensce. They are cited because it's considered unethical not to do so. The same would apply to using your source code.
Yeah. Also, if someone uses the code to do calculations they're going to publish, then the publication is going to need to include enough information so that readers will be able to determine how the calculations worked. The easiest and most straightforward way for to accomplish that is simply to give a reference to the paper describing the code. As a concrete example, when I was a grad student and postdoc doing nuclear physics, I used this program to do a lot of my calculations. Every time I published a paper that included those calculations, I needed to give a reference to T. Bengtsson, Nucl. Phys. A496, 56 (1989), because that was the only way to supply the reader with enough information to be able to understand (and possibly reproduce) exactly what I'd done.
Science works because we trust other scientists to cite our work if they use it. If we kept our work secret unless other scientists signed agreements to do so nothing would get done.
Yep. And the flip side of the ethical necessity for giving proper citations is that it's totally unethical to try to force someone to cite your work via some kind of artificial legal mechanism. Getting your work cited is the gold coin of academia, just as getting votes is the gold coin of politics. Forcing someone to cite your paper against their best judgment is as unethical as forcing someone to vote for you against their best judgment.
But the author is right on a point: it is a shame that a good article on Wikipedia can not be locked, as it is sure with time it will degrade as I have often found.
For the Robert A. Heinlein example, you can see that it was reviewed and listed as a featured article based on its state on June 11th, 2005. If you click on that date, it will bring you to a version of the article worthy of being used as a reference.
Three issues: (1) in academia, it's generally not appropriate to use any version of any encyclopedia article as a reference; (2) the vast majority of readers will only ever see the current version (and IMO the current version of the Heinlein example is much worse than the FA versions, especially in the lead); (3) the constant battle against entropy uses up a huge amount of energy from wikipedians, is no fun, and drives away people like me who used to enjoy editing on WP. I don't see #1 as a problem. If #2 and #3 are problems, then one solution is Citizendium.
My reading of that is that he wants to use the material in version B in his book/lecture. The original author wrote A. It has be changed in minor ways (less that 10%) by other people which is now called version B. He then wants to use this new version B because it has some information he must deem useful. He may be the sole author of version A and C (from your example), but if it is under the GFDL he cannot claim full rights to version B. If I understand the original premise and your counter-example.
I guess I can imagine three interpretations:
He wants to publish B in a book without even crediting the WP editors for their 10% contribution.
He wants to publish B in a book, with credit to the WP editors for their 10% contribution.
He wants to publish C in a book.
If it's 1, then he's a plagiarist.
If it's 2, then he's insane. That's just not going to fly in the academic world. I can just imagine this. He sends a cover letter, outline, and three sample chapters to an acquisitions editor at the University of Chicago Press. "Oh, by the way, I have 87 co-authors, known only by their fanciful pseudonyms on the Wikipedia web site. But their contributions were minor -- probably only 10% of the whole thing." No way. That type of academic publishing house is looking for original work from people who are leaders in their fields. They'd probably pass his letter around at their next meeting for laughs.
This is wrong for so many reasons that it's hard to know where to begin.
Okay, let's start with this quote:
"Say an academic (me) contributes a long article on London's Crystal Palace," [Russell Potter] wrote. "Others edit it in modest ways, but the article is still about 90% my own work. Perhaps I want to give a paper at a conference based on this entry, or use large bits of it in a book I'm writing. GFDL would have made either impossible."
This makes absolutely no sense. First off, if you're planning on publishing something as a book, the last thing on earth you want to do is put it on wikipedia, get it mixed promiscuously with writing by a hundred other people, and then cut and paste the result and send it to a publisher under your own name. That would be stupid -- so stupid that I assume it's not what Potter had in mind. I can only imagine that what he really had in mind was something more like this. He writes version A. He posts it on WP. On WP, it morphs into version B via other people's edits. Meanwhile, he decides he wants to write a book on it. He takes version A, edits it into a form that will work in the book, making version C. Now he's talking as if there's a licensing issue, but there's no licensing issue. He's the sole author and copyright holder of versions A and C. Licensing version B to WP under the GFDL doesn't even put him as the author under any obligations; licenses like the GPL and GFDL only impose obligations on other people, if they want to redistribute the material.
The idea about signing WP articles is pointless. First off, the whole culture of WP has always been based on the adrenaline rush of knowing that whatever edits you did, they'd immediately be visible to the whole world. That's why WP succeeded where Nupedia failed: instant gratification. Second, there is absolutely nothing stopping people from endorsing WP articles right now. Here you go, I, Ben Crowell, hereby endorse this version of the WP article on Robert Heinlein. (I think the current version is worse than that one.) What's that, you say? You say nobody cares that I endorsed that particular version, and the whole idea is boring and pointless? Well, yeah, I agree.
He hopes for "a 'gold rush' of users copying content from Wikipedia to Knol, where it can be verified by credentialed users and protected against vandalism." If he wants an encyclopedia where articles can be verified by credentialed users and protected against vandalism, it already exists: Citizendium. Citizendium's license is compatible with Wikipedia's. Why Knol? The only difference I can see is that Knol lets authors make money from ads placed next to their articles. But anyone who wants to make money from placing ads next to cut-and-pasted WP articles can already do that. It's perfectly legal, as long as they show the license.
I don't understand the point of a law prohibiting forged headers. DomainKeys/DKIM is a perfectly good technological solution to the problem of forged headers. Yahoo and Gmail both sign their outgoing mail with DK, and within the last few months Yahoo has gotten very aggressive about dropping incoming mail into the bitbucket if it doesn't have a DK signature and isn't coming from a domain that's on their whitelist. People running email servers are just going to have to bite the bullet and implement DK, unless they don't mind not having their users be able to send mail to Yahoo accounts.
But even though DK is rapidly bringing us to the point where forged headers are a thing of the past, that won't do a darn thing to end spam. Spammers can just open an account on Gmail, which is rapidly becoming the world'd biggest wellspring of both internet and usenet spam. The real benefit of DK is not that it ends spam, but that it saves organizations that sign their outgoing mail with DK from being blamed for spam that claims to be from them, but isn't.
So given that forged headers aren't the main problem, and forged headers can be fixed without legislation, I think the bar should be set very high for anyone proposing a legislative attack on forged headers. (But it does seem goofy to me to strike down such a law on free speech grounds. When spammers send me spam, they're using my resources without my permission. That's like putting their soapbox on my lawn instead of on the lawn in front of city hall.)
Okay, this is somewhat OT, but the article is on my short list for the award for worst web design. The link to page 2 doesn't function if you have javascript turned off. So okay, I told noscript to temporarily allow javascript on this page. But once you have javascript allowed, you get a distracting text banner scrolling across the top of the article, like something from someone's geocities homepage from 1995. Wow, let's combine the worst of web 1.0 and web 2.0!
Although it was demonstrably superior to other physics curricula, the PSSC program was ultimately a failure because publishers, who couldn't make much money selling the PSSC textbook due to competition, eventually dropped the book and pushed hard to get their proprietary, therefore more heavily marked-up, textbooks adopted by school boards.
I'm not sure I'd quite agree with that. I learned physics a decade or two after the PSSC era, and now teach physics. I agree that the PSSC books were of unusually high quality. However, they didn't go out of print. I believe Kendall-Hunt was bringing out new editions until very recently. A quick search on amazon turned up a 1995 edition by Houghton Mifflin. (Did Houghton Mifflin buy Kendall-Hunt or something?) I think the publishers customized the book with their own proprietary content as well. If you compare a Kendall-Hunt PSSC Physics book from the 1990's with one of the original ones from the 1960's, you might not even recognize them as the same book. I've heard a variety of reasons suggested as to why PSSC wasn't a smashing success. Arnold Arons (author of a well known book on physics pedagogy) thinks one factor was that the book made heavy use of reasoning involving ratios and proportionalities, which is difficult for many students, isn't taught in the K-12 math curriculum, and is something that even many high school physics teachers aren't comfortable doing. Another factor was almost certainly the unusual order of topics. The original PSSC text started with waves, and only got to Newton's laws many chapters later. If you look at the versions from 30 years later, I believe they all use a more traditional order of topics.
Thanks for the plug for Light and Matter -- I'm the author:-)
Their licensing scheme (CC-BY-SA) is compatible with mine (dual license, GFDL and CC-BY-SA), so if they want to adapt some of my materials, they can do that. My books are aimed at college classes, but I do have quite a few high school users. The problem with public high schools is that they usually have highly bureaucratic processes dictated by the state for selecting textbooks. (E.g., they want a sales rep from a big publisher to hold their hand and show them that the state standards say to cover Newton's first, second, and third laws, and -- lo and behold! -- their book covers Newton's first, second, and third laws. Some states also have rules about physical quality, etc.) For this reason, almost all of my adoptions from high schools have been from non-public schools, mainly Catholic schools.
By far, however, the absolute worst way to spend money on computer parts is to go to a big, corporate, well-lit box store where the guys have little name tags and same-color shirts.
It's always dangerous to generalize. I always buy parts from Fry's. The first couple of times I built my own machine, I botched some stuff. When I came back in with a woeful look on my face and walked up to the repair counter, they were just amazingly nice about it. The second time, the only problem had been that I hadn't pushed the cpu in far enough to seat it properly; they diagnosed the problem for me, and didn't charge me a dime. Okay, I suppose Fry's is qualitatively different from Circuit City.
But in terms of compatibility, deployment, and upgrades, they have the local app beat. [...] But they do answer issues of maintenance, upgrades, and control a lot better than locally installed apps.
I was with you for that whole first list, but you lost me when you got to "control." Control for whom?
But the reality is that the web app is the greatest advancement in maintenance since the mainframe/dumb-terminal.
In a way, web apps are a reversion to the mainframe/dumb-terminal model. You don't control what program you're running. Someone else does.
The big problem I have with web apps is that almost none of them are open-source. Just when I have thousands of debian packages worth of applications to choose from, why in the world would I want to revert to a model where half the code is client-side code that I as a user have no control over, and the other half is server-side code that I can't even see?
They also have "How To Think Like a Computer Scientist..." titles in C++ and Java. All the books are free as in speech. User-submitted reviews would be welcome at theassayer.org (make an account, search by title on the initial string "how to").
This seems thoroughly unsurprising. The higher the tax rate, the higher the incentive to cheat. Quebec has a sales tax rate of 12.875%, which is pretty high by south-of-the-border standards. The top marginal income tax rate in the U.S. from WWII until 1964 was 91%. Does anyone believe that rich people really paid 91% of their income to Uncle Sam? Of course not. They just hired people to find ways to avoid the tax. Action and reaction. Actually, Canada at least has made some efforts to harmonize their tax rates. If states in the U.S. wanted to increase the rate of collection of sales taxes, they would figure out ways of harmonizing their laws, and then it might be more practical to get rid of use tax, which is a joke, and charge the normal sales tax on interstate transactions. As it is, it's crazy. Every state may have dozens of different sales tax rates, and the list of taxable and nontaxable items is different in every state. For a small internet business with customers in all 50 states, it would be a prohibitive amount of work to pay taxes to all the states; you'd have to fill out 50 different annual tax forms, and calculate taxes on according to literally hundreds of local laws and rates. If they did that, they'd level the playing field, which currently treats bricks-and-mortar stores unfairly, and they'd also be able to lower their sales tax rates while still maintaining the same revenue.
They say prediction is difficult, especially about the future. Yahoo has a "political
dashboard" (flash app) that tries various things to predict the outcome of the presidential race. One technique they use is prediction markets, which are sort of similar to this thing about the wikipedia edits: instead of asking people their opinions, you watch their actions. In the yahoo dashboard app, you can click to switch between a map based on opinion polls and one based on prediction markets. One interesting thing is that the polls show Ohio leaning to McCain, but the prediction markets show it going to Obama. One thing that's really tough about predicting this election is that historically, racist white people have often lied to pollsters about their race-related opinions. Even though Obama is ahead in the polls, I'm kind of expecting that McCain will win, simply because the polls are likely to have this systematic error in them. OTOH, some people say that this racism-hiding effect in polls is no longer as strong as it used to be. The February Scientific American had an article that treated prediction markets with skepticism. Some of the evidence that people have been quoting in favor of prediction markets is apparently bogus, and nobody has the faintest clue how they really work.
It surprises me that more people haven't jumped over to Amazon. Maybe it's because of ignorance, I dunno... but you can get DRM-free MP3s there, encoded at 256kbit. When I use their download manager, and I have iTunes running, the song/album I downloaded automatically gets imported into iTunes.
Yeah, I don't understand why more people don't know about it. I guess Apple really did a good job of marketing the iPod and iTunes. Linux support on amazon's mp3 service is flaky, but here is a howto for linux users that explains how to get past some of the hassles.
[Perl] has the lowest defect rate of any open-source software product. [...] It has readily-accessible libraries for all types of programming tasks: Web application development, systems and network integration and management, end-user application development, middleware programming, REST and service-oriented architecture programming.
This essentially summarizes the reasons I prefer to use Perl: the quality of the implementation, and the good libraries. However, there is a dark side that we Perl lovers don't talk about much, which is that although Perl has good quality and good libraries, many of the libraries are not of good quality. My purpose here isn't to name names and rip into individuals who have contributed open-source code to CPAN out of the goodness of their hearts, but honestly, some of the code on CPAN is of very low quality and/or very poorly maintained. Quite a few CPAN libraries are basically glue that interfaces to some C code, and when you look at some of that C code, it looks like examples of the worst coding practices of the 1980's, before the internet existed, and before it really registered on coders' consciousnesses that buffer overflows, etc., were not just bugs but security holes. I've had a couple of bad experiences where I hitched my wagon to a particular CPAN module, and later had serious problems because that module was not actively maintained. E.g., crippling bugs would go unfixed for a year at a time.
On the other hand, I'm not sure that any of the other scripting languages come off any better. What the article says really is true: the base implementations of the other scripting languages are really not anywhere near as solid as Perl's is -- probably partly because Perl is so much older than the others, and therefore more mature. But this may change a lot in the future. Perl 6 is eventually going to be ready for prime time, and there will be a certain amount of chaos and confusion and bugginess at that point, as everyone adapts to the new environment. Also, Perl's head-start in terms of maturity will start to mean less and less as time goes on and the other scripting languages start to get more mature.
But... look at the data. That correlation is *terrible*. The phase is off.
IMO you're focusing on the wrong thing. If you believe their error bars, and believe that there are no systematic errors that correlate with time of year, then the probability that a correlation this strong would occur is quoted by them at 6*10^-18. On any plot like this, the human eye is a lousy judge of statistical quality. The statistical quality comes from the aggregate of all the data. You could have a plot like this with a thousand data points where a ruler line went through 2/3 of the error bars, and it could still be statistically inconsistent with zero modulation to an astronomical level of confidence.
The real weakness in this work is not in the statistical quality of the data, it's that it makes extraordinary claims, and extraordinay claims require extraordinary proof, but there are several obvious things that they should have done, and didn't:
They should have shown that the results were reproducible.
They should have shown that the results were reproducible under a variety of conditions that might have led to a bogus yearly modulation. For instance, they should have tried it in both the northern and the southern hemisphere, and they should have tried it in a temperature-controlled environment, with the temperature intentionally modulated with different phases.
They should have shown that the results occurred with qualitatively different types of detectors. For instance, they could use an isotope that emits both gammas and betas, and show that the same effect occurs both in an HPGe gamma-ray detector and in a silicon beta detector.
Although the statistical confidence level of the correlation is good, it's completely bogus that they didn't make any effort to improve the statistics so that the effect could be seen more clearly. If nuclear decay rates really depend on distance from the sun, then there's no way it's just going to be an effect that applies to some nuclei and not others. They have data from 32Si, which is an exotic isotope whose half-life is technically difficult to measure. Based on this extremely technically difficult measurement, they deduce an effect at the 10^-3 level. Well, with an isotope that's easier to produce, they could easily get their random errors down by several orders of magnitude; with a count rate of 1 kHz in each detector out of an array of 100, counting for 3 months, you'd have Poisson statistics of 10^-6. The error bars on the plot would then be too small to see. Basing the strongest possible claims on a design leading to the crappiest possible statistics is one of the hallmarks of junk science.
I also wouldn't discount small PCs even for having a Windows tax - I mean, I don't think you can buy Macs without buying the "OS X tax"?
The reason people refer to a "Windows tax" is that Windows is a monopoly, so you may not have a choice -- even if you're going to wipe Windows off it and install Linux, you may end up paying for the copy of Windows that came preinstalled on the machine. MS has fought tooth and nail to keep retailers from selling machines without an OS, machines with other OSes, and machines with multiple OSes.
I don't know what you were looking at, but this is the cheapest I could find on dell.com.
This was last week, with a discount advertised on their web site. Maybe it's not available now.
Nothing's changed, it's still apples to oranges. Just as you might not have needed built in sound then, you may not need FW, BT, dig audio, wireless net, IR remote, stack of dvds form factor now.
Sure, that's a valid point. The fact is that I personally don't want or need a single thing off of that list, but I'm sure other people do. I think the present era is really different, though, because the hardware features that differentiated a Mac from a PC used to be big, important features, whereas now they're really esoteric. Your analogy to the heated seats is a good one. When the Mac first came out, the alternative was a PC that displayed text using a character generator rather than a bitmapped screen; that was really a big deal, since it made the difference between being able to do desktop publishing and not being able to. I think a heck of a lot of people aren't really buying Macs for the added hardware features, or for the tight integration of hardware and OS -- they're actually buying them because it means they don't have to worry about viruses, or simply because they have a lot of the Mac UI programmed into their muscle memory (which is the reason I'll never be able to switch away from Linux -- I'm too used to having the Emacs keybindings universally available in essentially all the applications I use).
As far as the form factor, I agree that it's annoyingly hard to get a PC via a consumer retail channel with a small form factor, and it's ridiculous how they sell low-end PCs these days in tower cases that are 90% air, like Wonder Bread. But that's partly a matter of marketing -- people seem to associate a big case with a powerful machine. Small form factor PCs do exist, e.g., where I work most of the Windows boxes have the CPU integrated into a little compartment at the bottom of the LCD. It becomes more of an issue if you're trying to buy a PC to run Linux on without paying the Windows tax, because your options are narrow. But even this may be starting to change with, e.g., the Asus Eee Box, which is a desktop system (not a laptop) the size of a paperback book. Another thing to keep in mind is that if I buy the $280 Dell Linux box, I'm done, and I never have to pay any more money, whereas if I buy the $600 Mac Mini, the cost of the OS updates over the years is probably going to end up doubling the cost of the machine.
"The Mac at 25" makes me think back to when I bought my first mac in 1984. These days I'm on linux. My wife has an aging "iLamp" G4 on her desk, which we're probably going to get rid of soon and switch her to a linux box. But anyway I've continuously had a mac in the house for 24 years now.
Looking back, I see that time as dividing into three periods:
Early on, it was a revolutionary machine, way ahead of its time. The mouse, the GUI, and desktop publishing were all new. The price was high in 2008 dollars, but so was the price of an MS-DOS machine. It was a fairly open environment; you could buy the manual that described all the APIs ("Inside Mac") very cheaply, in a phone-book format. This was roughly the period of the 68000 cpu.
Then there was a period where it sucked more and more. By the time it got to MacOS 9, I was just finding it to be a completely untenable platform. Cooperative multitasking was a disaster, because it meant that anything that crashed was likely to crash your whole machine -- and the increasing complexity of the system and app software guaranteed that you'd have lots of crashes. There were tons of those little whatchamacallums -- were they called "extensions?" -- the little icons that showed up when you booted the box. The problem was that extensions would conflict or cause crashes. E.g., Adobe PageMaker would crash, and I'd call Adobe and ask if there was any way to avoid the crashes, etc., and they'd blame it on extension conflicts. So then I'd turn off every extension except for the Adobe extensions that were required to run PageMaker, and it would still crash. This was pretty much the PowerPC period. During this time, people would complain that macs were overpriced compared to PCs. That was kind of right and kind of wrong. It was wrong because it was an apples-to-oranges comparison. Macs came with lots of free hardware goodies, like sound I/O, that cheap PCs didn't. On the other hand, it was right, because if you didn't have the money for a Mac, you just didn't have a choice -- you were going to get a low-end PC, which was cheaper.
The third era is MacOS X. The big issue now is that low-end PCs can do everything I need, and low-end PCs are insanely cheap, so why buy a mac? E.g., if you price out a Dell PC with linux preinstalled, and omit the monitor, it's $249 for a dual-core 2 GHz, 2 Gb RAM, 250 Gb hd. The stereotype that was bogus in period #2 -- that macs were for people with too much money on their hands -- is really true now. It also really rankles me, as an early adopted of OS X, to think of how many of those $130 dot-upgrades I paid for, for several machines. One of the reasons we're dumping our last mac is that we stopped paying for the OS upgrades, which means the system is getting too old to get security updates, and new software (e.g., ff3) doesn't run.
Presumably to the manufacturers of hardware which contains binary-only drivers.
The idea is that it's a deliberately stress-testing distribution designed to be 100% Free and to cause any hardware which isn't Free to fail. If nobody complains that broken stuff is broken, it won't get fixed.
In practice, I'm having a really hard time believing that this is going to work. Customer: "I'm calling to complain that your driver doesn't work on my computer, because it isn't open source." Tech support guy: "What version of Windows are you running?" Customer: "I'm not running Windows, I'm running gNewSense." Tech support guy: "I'm sorry, we don't support gNewSense." Customer: "It's a version of Linux. Your web site says you support Linux." Tech: "Oh, Linux, cool. Yeah, we support Linux. I run Ubuntu at home myself. Yeah, it took a long time, but the higher-ups finally decided to support Linux. I can get you going, no sweat. Actually I'm surprised you had a problem at all. Our driver is in the mainline kernel and everything." Customer: "Ha, I know about your filthy driver. It has seventy-five bytes of hexadecimal in the source code that gets loaded into registers, and nobody knows what those 75 bytes do! It's unclean -- evil and unclean, I tell you! That's why I run gNewSense, which is purified of your nasty driver and with its insufficient level of freedom! Now please connect me with your CEO so I can show him the error of his ways!"
Might be a lot more effective to apply economic pressure by spreading the word about which hardware to buy that has good OSS driver support, rather than installing an entire distro designed to break your computer on the theory that breaking your own computer will make the manufacturer suffer.
Someone made a satirical comment elsewhere to this effect, but what good is free software if it mostly only supports Linux? I think using something like OpenBSD (is FreeBSD more open or is OpenBSD more free?) with only software that compiles natively on BSD is a true test of one's open and free nature.
I used to run FreeBSD on both my desktop and my server. (These days I'm running Ubuntu on my desktop, and Debian on the server.) My experience was that the vast majority of the apps you wanted were no problem at all -- just compile the port or install the binary package, no sweat. In the cases where there were problems, it was almost never a problem because the author of the ap used linuxisms; it more typically something like, e.g., I would upgrade library foo to a new version in order to satisfy make application bar work, but that would unexpectedly break application baz. In other words, it was a problem with the way the packaging had been done for BSD.
But I think I agree with the thrust of what you're saying. Different people have different ideas about what "free" means.
Slashdot Mirror
I agree with you 100% that Wikipedia has peaked. The quality of most articles is dropping over time, because anybody halfway sane doesn't want to pore autistically over a watchlist of cherished articles to make sure they don't succumb to entropy.
On the other hand, that doesn't mean that every dispute on WP is pointless, or that either side could be right on every issue. One bogus argument that's always posed by people who don't want their articles deleted is that it's not a paper encyclopedia, so there's no reason to keep the whole thing under a certain page count. Well, suppose Fred creates an article on his high school band, Fredsband, which only actually consisted of himself and his golden retriever. Every single time a user searches for "golden retriever," one of the hits is going to be the article on Fredsband. Also, when you have an article that's non-notable, it tends not to be linked to any other articles, and you get these little disjoint subsets of WP that are unhealthy. They can become havens for crackpots, or honeypots for spam links.
Here's a thought, use per machine accounts instead of per user accounts and click through the EULAs BEFORE imaging the drive.
That would break lots of other things we want (e.g., letting students read and write in their own networked folders), and it would be totally backwards to make decisions like that based on one piece of software.
I hope Sun will take a hint from this, and stop trying to impose the LGPL as a click-through EULA in the Windows versions of OpenOffice. I teach a physics lab course where I'm trying to encourage students to use OpenOffice instead of Office. (One really practical reason is that they'll make a graph using Excel at school, email it to themselves, then try to open it at home using Excel and find out they can't, because they have an older version of Excel at home.) The really annoying thing is that when you install OOo, it forces each user to click through the EULA the first time on that machine. This is lame, because:
Yeah. Also, if someone uses the code to do calculations they're going to publish, then the publication is going to need to include enough information so that readers will be able to determine how the calculations worked. The easiest and most straightforward way for to accomplish that is simply to give a reference to the paper describing the code. As a concrete example, when I was a grad student and postdoc doing nuclear physics, I used this program to do a lot of my calculations. Every time I published a paper that included those calculations, I needed to give a reference to T. Bengtsson, Nucl. Phys. A496, 56 (1989), because that was the only way to supply the reader with enough information to be able to understand (and possibly reproduce) exactly what I'd done.
Yep. And the flip side of the ethical necessity for giving proper citations is that it's totally unethical to try to force someone to cite your work via some kind of artificial legal mechanism. Getting your work cited is the gold coin of academia, just as getting votes is the gold coin of politics. Forcing someone to cite your paper against their best judgment is as unethical as forcing someone to vote for you against their best judgment.
Three issues: (1) in academia, it's generally not appropriate to use any version of any encyclopedia article as a reference; (2) the vast majority of readers will only ever see the current version (and IMO the current version of the Heinlein example is much worse than the FA versions, especially in the lead); (3) the constant battle against entropy uses up a huge amount of energy from wikipedians, is no fun, and drives away people like me who used to enjoy editing on WP. I don't see #1 as a problem. If #2 and #3 are problems, then one solution is Citizendium.
I guess I can imagine three interpretations:
If it's 1, then he's a plagiarist.
If it's 2, then he's insane. That's just not going to fly in the academic world. I can just imagine this. He sends a cover letter, outline, and three sample chapters to an acquisitions editor at the University of Chicago Press. "Oh, by the way, I have 87 co-authors, known only by their fanciful pseudonyms on the Wikipedia web site. But their contributions were minor -- probably only 10% of the whole thing." No way. That type of academic publishing house is looking for original work from people who are leaders in their fields. They'd probably pass his letter around at their next meeting for laughs.
If it's 3, then there are no legal issues at all.
This is wrong for so many reasons that it's hard to know where to begin.
Okay, let's start with this quote:
This makes absolutely no sense. First off, if you're planning on publishing something as a book, the last thing on earth you want to do is put it on wikipedia, get it mixed promiscuously with writing by a hundred other people, and then cut and paste the result and send it to a publisher under your own name. That would be stupid -- so stupid that I assume it's not what Potter had in mind. I can only imagine that what he really had in mind was something more like this. He writes version A. He posts it on WP. On WP, it morphs into version B via other people's edits. Meanwhile, he decides he wants to write a book on it. He takes version A, edits it into a form that will work in the book, making version C. Now he's talking as if there's a licensing issue, but there's no licensing issue. He's the sole author and copyright holder of versions A and C. Licensing version B to WP under the GFDL doesn't even put him as the author under any obligations; licenses like the GPL and GFDL only impose obligations on other people, if they want to redistribute the material.
The idea about signing WP articles is pointless. First off, the whole culture of WP has always been based on the adrenaline rush of knowing that whatever edits you did, they'd immediately be visible to the whole world. That's why WP succeeded where Nupedia failed: instant gratification. Second, there is absolutely nothing stopping people from endorsing WP articles right now. Here you go, I, Ben Crowell, hereby endorse this version of the WP article on Robert Heinlein. (I think the current version is worse than that one.) What's that, you say? You say nobody cares that I endorsed that particular version, and the whole idea is boring and pointless? Well, yeah, I agree.
He hopes for "a 'gold rush' of users copying content from Wikipedia to Knol, where it can be verified by credentialed users and protected against vandalism." If he wants an encyclopedia where articles can be verified by credentialed users and protected against vandalism, it already exists: Citizendium. Citizendium's license is compatible with Wikipedia's. Why Knol? The only difference I can see is that Knol lets authors make money from ads placed next to their articles. But anyone who wants to make money from placing ads next to cut-and-pasted WP articles can already do that. It's perfectly legal, as long as they show the license.
I don't understand the point of a law prohibiting forged headers. DomainKeys/DKIM is a perfectly good technological solution to the problem of forged headers. Yahoo and Gmail both sign their outgoing mail with DK, and within the last few months Yahoo has gotten very aggressive about dropping incoming mail into the bitbucket if it doesn't have a DK signature and isn't coming from a domain that's on their whitelist. People running email servers are just going to have to bite the bullet and implement DK, unless they don't mind not having their users be able to send mail to Yahoo accounts.
But even though DK is rapidly bringing us to the point where forged headers are a thing of the past, that won't do a darn thing to end spam. Spammers can just open an account on Gmail, which is rapidly becoming the world'd biggest wellspring of both internet and usenet spam. The real benefit of DK is not that it ends spam, but that it saves organizations that sign their outgoing mail with DK from being blamed for spam that claims to be from them, but isn't.
So given that forged headers aren't the main problem, and forged headers can be fixed without legislation, I think the bar should be set very high for anyone proposing a legislative attack on forged headers. (But it does seem goofy to me to strike down such a law on free speech grounds. When spammers send me spam, they're using my resources without my permission. That's like putting their soapbox on my lawn instead of on the lawn in front of city hall.)
Okay, this is somewhat OT, but the article is on my short list for the award for worst web design. The link to page 2 doesn't function if you have javascript turned off. So okay, I told noscript to temporarily allow javascript on this page. But once you have javascript allowed, you get a distracting text banner scrolling across the top of the article, like something from someone's geocities homepage from 1995. Wow, let's combine the worst of web 1.0 and web 2.0!
I'm not sure I'd quite agree with that. I learned physics a decade or two after the PSSC era, and now teach physics. I agree that the PSSC books were of unusually high quality. However, they didn't go out of print. I believe Kendall-Hunt was bringing out new editions until very recently. A quick search on amazon turned up a 1995 edition by Houghton Mifflin. (Did Houghton Mifflin buy Kendall-Hunt or something?) I think the publishers customized the book with their own proprietary content as well. If you compare a Kendall-Hunt PSSC Physics book from the 1990's with one of the original ones from the 1960's, you might not even recognize them as the same book. I've heard a variety of reasons suggested as to why PSSC wasn't a smashing success. Arnold Arons (author of a well known book on physics pedagogy) thinks one factor was that the book made heavy use of reasoning involving ratios and proportionalities, which is difficult for many students, isn't taught in the K-12 math curriculum, and is something that even many high school physics teachers aren't comfortable doing. Another factor was almost certainly the unusual order of topics. The original PSSC text started with waves, and only got to Newton's laws many chapters later. If you look at the versions from 30 years later, I believe they all use a more traditional order of topics.
Thanks for the plug for Light and Matter -- I'm the author :-)
Their licensing scheme (CC-BY-SA) is compatible with mine (dual license, GFDL and CC-BY-SA), so if they want to adapt some of my materials, they can do that. My books are aimed at college classes, but I do have quite a few high school users. The problem with public high schools is that they usually have highly bureaucratic processes dictated by the state for selecting textbooks. (E.g., they want a sales rep from a big publisher to hold their hand and show them that the state standards say to cover Newton's first, second, and third laws, and -- lo and behold! -- their book covers Newton's first, second, and third laws. Some states also have rules about physical quality, etc.) For this reason, almost all of my adoptions from high schools have been from non-public schools, mainly Catholic schools.
It's always dangerous to generalize. I always buy parts from Fry's. The first couple of times I built my own machine, I botched some stuff. When I came back in with a woeful look on my face and walked up to the repair counter, they were just amazingly nice about it. The second time, the only problem had been that I hadn't pushed the cpu in far enough to seat it properly; they diagnosed the problem for me, and didn't charge me a dime. Okay, I suppose Fry's is qualitatively different from Circuit City.
I was with you for that whole first list, but you lost me when you got to "control." Control for whom?
In a way, web apps are a reversion to the mainframe/dumb-terminal model. You don't control what program you're running. Someone else does.
The big problem I have with web apps is that almost none of them are open-source. Just when I have thousands of debian packages worth of applications to choose from, why in the world would I want to revert to a model where half the code is client-side code that I as a user have no control over, and the other half is server-side code that I can't even see?
They also have "How To Think Like a Computer Scientist..." titles in C++ and Java. All the books are free as in speech. User-submitted reviews would be welcome at theassayer.org (make an account, search by title on the initial string "how to").
Currently the retailer does need to pay sales taxes to any state where they have a physical presence, or "nexus."
Use tax is paid at the rate of the purchaser's home state.
If you have a "nexus" in the customer's state, you pay the rate in the customer's state.
This seems thoroughly unsurprising. The higher the tax rate, the higher the incentive to cheat. Quebec has a sales tax rate of 12.875%, which is pretty high by south-of-the-border standards. The top marginal income tax rate in the U.S. from WWII until 1964 was 91%. Does anyone believe that rich people really paid 91% of their income to Uncle Sam? Of course not. They just hired people to find ways to avoid the tax. Action and reaction. Actually, Canada at least has made some efforts to harmonize their tax rates. If states in the U.S. wanted to increase the rate of collection of sales taxes, they would figure out ways of harmonizing their laws, and then it might be more practical to get rid of use tax, which is a joke, and charge the normal sales tax on interstate transactions. As it is, it's crazy. Every state may have dozens of different sales tax rates, and the list of taxable and nontaxable items is different in every state. For a small internet business with customers in all 50 states, it would be a prohibitive amount of work to pay taxes to all the states; you'd have to fill out 50 different annual tax forms, and calculate taxes on according to literally hundreds of local laws and rates. If they did that, they'd level the playing field, which currently treats bricks-and-mortar stores unfairly, and they'd also be able to lower their sales tax rates while still maintaining the same revenue.
They say prediction is difficult, especially about the future. Yahoo has a "political dashboard" (flash app) that tries various things to predict the outcome of the presidential race. One technique they use is prediction markets, which are sort of similar to this thing about the wikipedia edits: instead of asking people their opinions, you watch their actions. In the yahoo dashboard app, you can click to switch between a map based on opinion polls and one based on prediction markets. One interesting thing is that the polls show Ohio leaning to McCain, but the prediction markets show it going to Obama. One thing that's really tough about predicting this election is that historically, racist white people have often lied to pollsters about their race-related opinions. Even though Obama is ahead in the polls, I'm kind of expecting that McCain will win, simply because the polls are likely to have this systematic error in them. OTOH, some people say that this racism-hiding effect in polls is no longer as strong as it used to be. The February Scientific American had an article that treated prediction markets with skepticism. Some of the evidence that people have been quoting in favor of prediction markets is apparently bogus, and nobody has the faintest clue how they really work.
Yeah, I don't understand why more people don't know about it. I guess Apple really did a good job of marketing the iPod and iTunes. Linux support on amazon's mp3 service is flaky, but here is a howto for linux users that explains how to get past some of the hassles.
This essentially summarizes the reasons I prefer to use Perl: the quality of the implementation, and the good libraries. However, there is a dark side that we Perl lovers don't talk about much, which is that although Perl has good quality and good libraries, many of the libraries are not of good quality. My purpose here isn't to name names and rip into individuals who have contributed open-source code to CPAN out of the goodness of their hearts, but honestly, some of the code on CPAN is of very low quality and/or very poorly maintained. Quite a few CPAN libraries are basically glue that interfaces to some C code, and when you look at some of that C code, it looks like examples of the worst coding practices of the 1980's, before the internet existed, and before it really registered on coders' consciousnesses that buffer overflows, etc., were not just bugs but security holes. I've had a couple of bad experiences where I hitched my wagon to a particular CPAN module, and later had serious problems because that module was not actively maintained. E.g., crippling bugs would go unfixed for a year at a time.
On the other hand, I'm not sure that any of the other scripting languages come off any better. What the article says really is true: the base implementations of the other scripting languages are really not anywhere near as solid as Perl's is -- probably partly because Perl is so much older than the others, and therefore more mature. But this may change a lot in the future. Perl 6 is eventually going to be ready for prime time, and there will be a certain amount of chaos and confusion and bugginess at that point, as everyone adapts to the new environment. Also, Perl's head-start in terms of maturity will start to mean less and less as time goes on and the other scripting languages start to get more mature.
IMO you're focusing on the wrong thing. If you believe their error bars, and believe that there are no systematic errors that correlate with time of year, then the probability that a correlation this strong would occur is quoted by them at 6*10^-18. On any plot like this, the human eye is a lousy judge of statistical quality. The statistical quality comes from the aggregate of all the data. You could have a plot like this with a thousand data points where a ruler line went through 2/3 of the error bars, and it could still be statistically inconsistent with zero modulation to an astronomical level of confidence.
The real weakness in this work is not in the statistical quality of the data, it's that it makes extraordinary claims, and extraordinay claims require extraordinary proof, but there are several obvious things that they should have done, and didn't:
The reason people refer to a "Windows tax" is that Windows is a monopoly, so you may not have a choice -- even if you're going to wipe Windows off it and install Linux, you may end up paying for the copy of Windows that came preinstalled on the machine. MS has fought tooth and nail to keep retailers from selling machines without an OS, machines with other OSes, and machines with multiple OSes.
Thanks, that's a very informative post!
This was last week, with a discount advertised on their web site. Maybe it's not available now.
Sure, that's a valid point. The fact is that I personally don't want or need a single thing off of that list, but I'm sure other people do. I think the present era is really different, though, because the hardware features that differentiated a Mac from a PC used to be big, important features, whereas now they're really esoteric. Your analogy to the heated seats is a good one. When the Mac first came out, the alternative was a PC that displayed text using a character generator rather than a bitmapped screen; that was really a big deal, since it made the difference between being able to do desktop publishing and not being able to. I think a heck of a lot of people aren't really buying Macs for the added hardware features, or for the tight integration of hardware and OS -- they're actually buying them because it means they don't have to worry about viruses, or simply because they have a lot of the Mac UI programmed into their muscle memory (which is the reason I'll never be able to switch away from Linux -- I'm too used to having the Emacs keybindings universally available in essentially all the applications I use).
As far as the form factor, I agree that it's annoyingly hard to get a PC via a consumer retail channel with a small form factor, and it's ridiculous how they sell low-end PCs these days in tower cases that are 90% air, like Wonder Bread. But that's partly a matter of marketing -- people seem to associate a big case with a powerful machine. Small form factor PCs do exist, e.g., where I work most of the Windows boxes have the CPU integrated into a little compartment at the bottom of the LCD. It becomes more of an issue if you're trying to buy a PC to run Linux on without paying the Windows tax, because your options are narrow. But even this may be starting to change with, e.g., the Asus Eee Box, which is a desktop system (not a laptop) the size of a paperback book. Another thing to keep in mind is that if I buy the $280 Dell Linux box, I'm done, and I never have to pay any more money, whereas if I buy the $600 Mac Mini, the cost of the OS updates over the years is probably going to end up doubling the cost of the machine.
"The Mac at 25" makes me think back to when I bought my first mac in 1984. These days I'm on linux. My wife has an aging "iLamp" G4 on her desk, which we're probably going to get rid of soon and switch her to a linux box. But anyway I've continuously had a mac in the house for 24 years now.
Looking back, I see that time as dividing into three periods:
In practice, I'm having a really hard time believing that this is going to work. Customer: "I'm calling to complain that your driver doesn't work on my computer, because it isn't open source." Tech support guy: "What version of Windows are you running?" Customer: "I'm not running Windows, I'm running gNewSense." Tech support guy: "I'm sorry, we don't support gNewSense." Customer: "It's a version of Linux. Your web site says you support Linux." Tech: "Oh, Linux, cool. Yeah, we support Linux. I run Ubuntu at home myself. Yeah, it took a long time, but the higher-ups finally decided to support Linux. I can get you going, no sweat. Actually I'm surprised you had a problem at all. Our driver is in the mainline kernel and everything." Customer: "Ha, I know about your filthy driver. It has seventy-five bytes of hexadecimal in the source code that gets loaded into registers, and nobody knows what those 75 bytes do! It's unclean -- evil and unclean, I tell you! That's why I run gNewSense, which is purified of your nasty driver and with its insufficient level of freedom! Now please connect me with your CEO so I can show him the error of his ways!"
Might be a lot more effective to apply economic pressure by spreading the word about which hardware to buy that has good OSS driver support, rather than installing an entire distro designed to break your computer on the theory that breaking your own computer will make the manufacturer suffer.
I used to run FreeBSD on both my desktop and my server. (These days I'm running Ubuntu on my desktop, and Debian on the server.) My experience was that the vast majority of the apps you wanted were no problem at all -- just compile the port or install the binary package, no sweat. In the cases where there were problems, it was almost never a problem because the author of the ap used linuxisms; it more typically something like, e.g., I would upgrade library foo to a new version in order to satisfy make application bar work, but that would unexpectedly break application baz. In other words, it was a problem with the way the packaging had been done for BSD.
But I think I agree with the thrust of what you're saying. Different people have different ideas about what "free" means.