It's surely true that the vast majority of parents in the US wouldn't want their children to see what you described. Whether that would be true in all cultures is open to question. What's a web site provider in a culture that considers something like that normal to do?
That notwithstanding, you're also assuming that all web site providers are going to rate their sites in good faith and in accordance with your morals. Neither of these is going to be consistently true. There will be sites that intentionally misrate themselves, and there will be others that will be rated according to very different standards. The ratings on those sites will not even be useful as a starting point. Indeed, they will be worse, because they'll lull you into a false sense of security.
Part of my posting was explaining the DIFFERENCE between movie ratings and web site ratings. Movie ratings work in their own somewhat perverted way for several reasons:
1) There are relatively few movies made, by relatively few organizations, that achieve wide distribution.
2) There is a central ratings organization (in the US) that applies the ratings. This is possible because of the relatively few movies made, but also because relatively few places exhibit movies and they all have much the same policies.
3) The central ratings organization only rates movies for the US (other countries have their own boards), which is fairly homogeneous culturally (compared to the world at large).
4) Movies are static content (they don't change constantly -- if you see Titanic today, it's the same movie you saw a year ago).
NONE of these conditions are true for the web! There are probably on the order of six orders of magnitude more web sites than there are movies made in any one year. That's a factor of a million. It is not possible for one organization to "screen" all of these sites, so ratings will have to be applied by the site operators. Many of these sites are not located in the United States, so ratings will be applied according to the standards of the operators of those sites. And finally, sites change constantly.
You then have to take into account that there's no organized web site industry that can enforce this kind of cartel, as each web site operator is independent of every other operator. There are web site providers who are malicious -- who WANT to expose your children to pornography. What's going to happen when one of these sites is rated G? Are you going to relax your guard, since the G rating means that it's good for your children, or will you still have to prescreen it against these malicious operators?
And before you say that these folks are breaking ze law and must be put in prison, you first have to find them. Tracking these folks down is hard, especially when they live and provide service overseas. And speaking of overseas, what about people overseas who simply have different values from you and think that full frontal nudity is a normal part of life and that there's no reason children should not be exposed to it?
My point is that this "enforced ratings system worldwide" simply cannot work, so there's no use talking about any potential benefits or uses of it. It doesn't matter what you think the potential benefits may or may not be; since the system cannot possibly work as you intend, these benefits don't exist.
The ONLY way that filtering can work is on the client side. Whether you do it personally (which has the highest accuracy level) or you use software produced by someone you trust, or you rely on voluntary ratings (which can be subverted unless there's some real authentication behind it, which still requires what amounts to client-side verification), THIS MUST HAPPEN AT YOUR END! You CANNOT rely on the server to give you ANY useful information at all about appropriateness of the content. Period.
Now do you understand my point? What you're asking for cannot exist, so the whole basis for your argument is flawed.
You haven't answered any of my objections, merely reiterated your original message. There's absolutely no way that "mandatory ratings" can be enforced; content providers can always disagree with your interpretation (intentionally or otherwise); and how does this help parents who have different ideas about what matter?
Remember that this is very different from the movie industry, where everyone's cooperating (many would say "colluding" or "conspiring") to make the rating system "work" after a fashion, or television, which is basically controlled by a few large corporations (and even with cable there are only a relatively few discrete channels). On the internet, there are MILLIONS of content providers who really don't care about your personal values, and some who would probably like to actively subvert them. This kind of thing simply won't work on the net.
If you really want to CONTROL what your children see on the net, simply arrange to be there at all times with them. There's no other way to do it. If that doesn't work, learn to live with the fact that the world is imperfect and find a better way to deal with it. There's no more gentle way of putting it.
It's bad enough that movies are de facto mandated ratings (theaters won't show unrated or NC-17 movies). As for video games and comics, are they rated (in the US, at least)? I wasn't aware of that. If people WANT to rate themselves, that's fine. If not, it's another matter.
One obvious problem with mandatory ratings of web sites is: who's going to do the rating? There are so many web sites and they change on such a continuous basis that there's no possible way that a centralized rating authority can assign ratings to every web page (or really every separately addressable object) in existence. That means that content providers have to supply their own ratings. There's obviously a conflict of interest here. The Australian law seems to try to get around that by requiring that the ratings coincide with what the central censorship bureau would assign the site. That makes it difficult, to say the least, for people to safely rate their own sites. This issue has been discussed in a lot more depth elsewhere, and I don't feel like a lot of typing.
As for this issue of parents vis a vis children, if you want your children to be that coddled, then you can bloody well find the time to research every single thing your child wants to buy/see/use. If you don't want to, you can buy filtering software to do the job for you. If you complain that it doesn't work well enough, then write better software yourself to do it. But don't ask the person running the erotica site to admit that his material is morally wrong and to do your work for you. If nothing else, it won't work. Mandating that every web site be correctly rated is impossible. Don't pretend that it will work -- it won't, and your child will be exposed to all the "garbage" (distinctly a value judgment) that he'd be exposed to anyway. If you think that "mandatory" ratings will relieve you of any effort, you're fooling yourself, I'm afraid.
Besides which, ratings are a very limited way of measuring something. I presume you're thinking about ratings for erotic and violent contact. But what about an Islamic parent who doesn't want his son to see pictures of women with their legs uncovered? Or a Jewish parent who doesn't want her children to get all caught up in Christmas? How does any rating system "protect" them?
Isn't it better to teach your children how to properly handle material that you or they consider undesirable than to pretend that you can magically stop it?
I've been hacking on a driver for the Epson Stylus Photo EX for a while. This is actually Michael Sweet's driver for the Gimp print plugin, ironically enough.
It turns out that Epson has (at least) two levels of their developer documentation, beyond the standard documentation for their printer language (ESCP/2). The "Level 1" documentation on their developer web site (http://www.ercipd.com/isv/edr_docs.htm) gives fairly detailed documentation on the printer, but there's at least one mroe level of documentation that's available only under NDA, which (obviously) I don't have. So I've been able to get very good print quality, but not the level of quality the printer is ultimately capable of. I've found another Ghostscript driver around that apparently does know how to achieve the higher print quality (looking carefully at the printer output it's obvious -- with an 8x loupe and simply looking at the micro-banding, it's clearly doing something different), but I haven't had time to study the code yet (my driver seems to produce better fidelity).
BTW, if anyone would like to hack on my modifications to Michael's Gimp plugin, it's available at http://www.tiac.net/users/rlk/print.tar.gz. I'd like to find more people to test this and make suggestions; it's become a rather expensive hack (in terms of paper, and particularly ink).
Whatever my thoughts about the multiple levels of documentation, though, I don't particularly blame Epson for putting the smarts in a software driver. Dithering algorithms and such for 6 colors (CMYK + light C + light M) seem to become very complex very quickly, and having this stuff be under software control makes it much easier to experiment with changes. For example, my driver achieves higher color fidelity (visually compared to my monitor) than the driver Epson bundles. It actually makes the printer a MORE open device than if the firmware were embedded. I wish Epson would put the entire programming information on its web site, not just parts of it, but at least I (or others who know more about color and dithering than I do) have a chance of reverse engineering the format and improving the output quality beyond what even Epson achieves. A true PostScript printer certainly has a very open interface, but a closed implementation. As much CPU as this consumes, I think that for printers such as this, having an open implementation is very valuable.
Examples of improvements that could be made:
1) More user-level controls (I've added gamma, contrast, and color level controls to the existing set). This enables fine tuning for a particular image or monitor.
2) Different kinds of paper. Images print very differently on different papers. Photo-quality glossy paper is different from "photo-quality" coated paper, which in turn is much, much different from standard ink-jet paper. This is mostly due to how much ink the paper absorbs. People who actually know what they're doing here (as opposed to dabblers such as myself who like to shoot off our mouths:-) ) could tune the output for different papers.
3) Different inks. I bought a third party cartridge because it was the only one I could find at CompUSA. Its color balance was somewhat different. Conceivably a third party might devise archival inks (that don't fade rapidly) that have different colors and densities. With a software driver the printer really does become open; if it's in ROM on the printer, there's no adjustment for it.
4) Improved dithering algorithms. Someone might devise better dither patterns to get smoother output.
Red Hat did exactly what you claim Debian did (I don't actually know what Debian did) -- they allow use of Red Hat in the name (e. g. "derived from Red Hat Linux" or whatnot), as long as it's clear that it's not the official package. I don't have the URL to the letter that they sent out about it, but it got heavily publicized on Slashdot and Linux Today, among others.
Red Hat acted honorably here. They want to use the trademark to make it clear to customers what they're getting, not to screw anybody (except people who wanted to pretend that they were selling the official package with support, as opposed to a cloned CD).
It simply means that an institution of higher learning has chosen to grant somebody a high honor for whatever reason of their choice. Some places (e. g. MIT) don't grant them at all; others (such as the little red brick community college up the creek, sometimes known as Harvard) do.
I don't believe that it entitles him to use "Dr." as a title. So in some ways it means more, and in more ways it means less, than a true PhD. But it certainly indicates that a (in this case prestigious) institution of higher learning has seen fit to bestow a major honor on him. To reject it (as a few people suggested) would be a major insult, I should think.
Linus has certainly done a lot of good for a lot of people. See, for example, the project in Mexico placing computers running Linux in classrooms -- Microsoft software is simply too expensive to permit that. One can argue about how much code Linus has written vs. the FSF vs. Berkeley vs. the other kernel hackers, but there's little question in my mind that he did considerably more than just be in the right place at the right time and go along for the ride.
One will note that Linus's message announcing 2.3.18 merely indicated that he would be away for a few weeks and not reading email. It implied nothing more than a well-deserved vacation after having feature frozen 2.4. He made absolutely no mention of the honorary doctorate.
Somehow, this doesn't strike me as the mark of a pompous ass.
My father was a chemical engineer. He did have a PE certification in a couple of states, but it very seldom mattered during his employment. After he retired and went into consulting, though, it did matter for certain jobs, for exactly the reasons specified -- a lot of his work involved investigations after accidents and the like, and he frequently had to write reports that would have legal standing in the event that there were any lawsuits over, say, denied insurance claims.
My father was only licensed to practice in two states. As it happens, there are only a few states that matter from the standpoint of the particular subspecialty of chemical engineering that he practiced.
Chemical engineering is a very traditional field where pieces of paper mean a lot and (at least 15 years ago) everyone wore suits to work every day. Even so, it sounded like his PE really only mattered very much when he was consulting.
The PE examination requires substantial knowledge in many areas -- electrical, chemical, civil, and such. For the vast majority of people in the software (and even hardware) field who style themselves (or are styled by their companies) "engineer", this has no relevance whatsoever. I, for example, mostly do OS and other low level work. I'd be surprised if there are more than a handful of people in the building who even know what a PE is, much less what one does.
Most software vendors seem to subscribe to the "we don't accept liability" school of thought. NT's EULA, which apparently states that NT is not certified for use in safety-critical situations, is a case in point. If they get sued over somebody's death because of an OS failure, they're simply going to point to the EULA and state that they made it quite clear that this is not qualified for the purpose. Of course, any PE at the nuclear power plant who signed off on using such a product would probably be in trouble, but any software vendor that takes that kind of approach probably doesn't need a PE on staff.
EE's who design consumer electronics probably don't need a PE at all. Most contractors probably don't, either. However, if you want contract jobs in situations where you might be called to testify in court (such as investigations of problems after the fact), or where you have to certify that something was done according to best accepted practice, it's probably essential. So companies that write custom software to run nuclear power plants probably do need some people with PE certification.
I wrote a letter of nomination for Alan before I read this discussion. Enclosed forthwith.
I hereby nominate Alan Cox for the 1999 Free Software Award nomination.
Alan Cox is perhaps Linus Torvalds's chief lieutenant. I think Alan should be recognized, however, for a very specific task that he has willingly taken on: maintenance of old stable releases of the Linux kernel. This is not a glamorous task, but is one that is essential to counter the impression that free software lacks support.
Alan has taken it upon himself to maintain the 2.0 and 2.2 Linux kernels while others forge ahead with the more exciting work of designing new features for the development kernels. While the stability of Linux 2.0 is legendary, bugs are occasionally discovered, new devices become available, and security holes are found. In order to be accepted for mission critical applications, old versions of software must be supported, with critical items such as these fixed. Responsible software vendors have always provided such support, and Alan's work has made it possible for free software to be held to such exacting standards.
In addition to this key responsibility, Alan is also a major contributor to the development kernel, and particularly concerns himself with low level kernel functionality and devices, which are notoriously complex yet essential pieces of functionality.
Finally, Alan ceaselessly promotes free software in all forums in which he participates.
For these reasons, I believe that Alan Cox would be a deserving recipient of the 1999 Free Software Award.
Actually, even in the heart of the tornado belt, any one location is struck on average once every 2000 years. The repeat frequency for large hail or destructive wind is, of course, much more frequent. However, the important point is that there's no location within that zone that's particularly more or less prone to devastating effects.
Residents of barrier islands and the immediate coast in hurricane country (the entire east coast, in other words) are in quite a different situation. Significant storm surge isn't as rare as that, and what's more, there are well-understood reasons why those locations are much more vulnerable than even a few miles inland. Building somewhere that any competent meteorologist or structural engineer can tell you is a sitting duck for a fairly common kind of storm is asking for trouble.
Before people talk about Homestead and Andrew, let me point out that Homestead was really at ground zero for a particularly devastating storm; that quite a few well-constructed buildings did survive; and that there was evidence of a previously unknown phenomenon (extreme wind gusts resembling microbursts in the eyewall) that is hard to defend against by any means, and that really is very localized and random.
Storm surge affects a much greater frontage of shore, and is far more devastating than even a hurricane's rain and wind. There's also a simple defense against it -- just don't build anything close to certain kinds of shoreline that you're not willing to have destroyed every so often. Nothing will really stop a storm surge other than distance. Winter storms also cause destructive surges from Cape Hatteras north, so there's no excuse for excessive development too close to the northern shoreline (there's also less of it, because it's less desirable, and because winter storms are much more common).
It's not only hurricanes that are an issue -- anyone who builds a house in a known floodplain, or close to a volcano that might be active, or in an obvious mudslide channel is also being pretty silly, when it's not hard to choose a much safer location nearby.
Meanwhile the storm is still tracking to the left of the forecasts. Check out ne.weather and sci.geo.meteorology for a lot of discussion on it. It's starting to turn (seemingly fairly decisively finally), but it's going to be an awfully close shave at best.
Probably very few people here (I number among the "many") have ever seen a major hurricane make a close pass. Here in the northeast, the only hurricanes I've seen were David (1979), Gloria (1985), and Bob (1991). Each of these were fairly weak by the time they reached us. If Floyd does come anywhere close to the location of the webcam, there will be a lot of opportunities to see what this kind of storm can really do.
The truth of the matter is that a major hurricane can hit anywhere along the east coast and the Gulf of Mexico, and in the northeast, if a major hurricane does hit, it can carry strong winds a considerable distance inland (because such a storm is invariably moving very fast -- 40 MPH forward speed or greater -- and simply doesn't have time to weaken, and furthermore the right hand side of such a storm has its winds enhanced by the forward motion of the storm). Hazel in 1954 went inland at South Carolina and screamed north at 50 MPH or so, so it had been inland for 500 miles or thereabouts by the time it reached New York's latitude. Nonetheless, it spawned a wind gust of over 100 MPH at the official observing station (probably Central Park), which was well east of the center. The 1938 hurricane was cat 3 (still a major hurricane) when it reached the coast, and Blue Hill had a peak wind (1 minute, which is not a gust) of 183 MPH in that storm!
Anyway, this storm might nail (to some degree) the entire east coast. If the storm does hit hard, and if the webcam survives, it might help warn other people what to really expect from a major storm. It's weakening a bit (down to 125 knots from 135 last night), but not enough to matter. The one piece of good news for Florida is that if it is a coast hugger the coast will be hit with the less severe west side of the storm. However, it's an absolutely gigantic system, so even a moderately near miss will cause problems. It's not a tiny system like Andrew was.
I'm also on the "have a core of employees for core product" side, for reasons that people have elucidated. While I agree that even employees should not assume absolute job security, I don't believe that most companies are going to take that short term of a horizon toward their employees. And in some ways, if things turn bad an employee is in a better position to up and leave than a contractor. Leaving in the middle of a contract doesn't strike me as a terribly wonderful thing to do, whereas employees are always free to leave whenever they please.
Furthermore, this attitude "everyone's short term anyway, so I might as well be a contractor, rake in the $$$, and keep jumping aroune" is exactly wrong for a company trying to build up a long term base of experience (which should be any company trying to build a product, as opposed to a company that's merely a body shop). At my previous company, we scrutinized very carefully anybody who seemed to jump around a lot. Our CEO firmly believed in employees rather than contractors for all but very clearly defined, very specific projects, and we all agreed with him on that issue.
I can think of one particular exception at an earlier job, though. This was an individual who wanted a contract specifically for benefits reasons -- he had some medical condition that was well covered by his insurance, and he didn't want to risk having any coverage problems being on employee health benefits and preferred the extra cash to useless health insurance. However, it was understood that he was in all other respects an employee. That worked out, but the important thing is that he had an employee attitude, and wished to contract for unique reasons related to benefits.
Under no circumstances would I hire a contractor as a lead system administrator. I'm certain that there are plenty of them who do conscientious work, and I'm well aware that there are disgruntled employees, I would much prefer to have a system administrator whose incentives and mindset are structured toward a long-term relationship with the company. Certainly I would permit hiring a contractor for specific tasks, such as assisting in a major network expansion or reconfiguration, but the person who's in charge of systems and security is even more important to keep around and suitably gruntled than, say, a VP of engineering.
Anybody who tries to compare any PC to a Starfire is obviously going to look silly -- if you're running a workload that the Starfire can run effectively (nicely parallel). But then again, throw together a Beowulf cluster with enough cheap PC's and run something that's reasonably close to embarrassingly parallel, and you're right up there in the list.
Somehow, it seems awfully cynical to call anything that's in the 3-digit mips and Mflops range a "puny toy". Any of these things has the raw CPU to blow a Cray-1 out of the water (of course, it doesn't have the memory bandwidth, but that's another matter). It's also a very large fraction of what we could do on a Connection Machine (CM-2) 10 years ago. Sure, I'm well aware that I can spend less on a processor than on the latest Star Wars toy for my nephew, and likewise that processor technology's advancing at an incredible rate, but these chips would have been called supercomputers not too many years ago.
And then I shudder to think how these things actually slow down as much as they do running Windows, and I can't imagine what must be going on that mere UI display (without any speech recognition) can bog one of these things down. Oh well, I guess I can just be smug about the fact that I can't even *run* Windows any more, since my last mobo upgrade...
Computernerd.com isn't pretending that their Celerons are PII's, so what fraud is involved? As far as 300/450A's vs. 466's, if you really can clock a 466 to 700 MHz, then yes, it will beat the 300/450, but unless you can do that, the 450 will win anything that goes significantly out of cache (just about everything except a synthetic computation kernel). Why? Because Celerons are overclocked by boosting the FSB frequency from 66 MHz to something else (100 for the 300/450, for example). This reduces the memory bottleneck substantially.
Certainly it's a firing offense anywhere (and perhaps grounds for a defamation lawsuit). The question is whether Microsoft actually wants to go after this individual or not.
Notice that:
1) Thus far they have been unable to identify the employee in question (which if their gateway is even halfway decent shouldn't take more than a few hours at most).
2) Rob Bennett (director of marketing, MSN) didn't exactly condemn whoever did it "somebody who got a little over-passionate but went about it the wrong way." Sounds like praising with faint damnation to me, especially in a culture that rewards taking risks.
3) Not only does Bennett not condemn the unnamed employee, he concurs with the statements so made.
It probably (IMHO) was not officially sanctioned, but I suspect that there was no need to actually ask someone to do so or formally approve it.
There needs to be a balance between fixing bugs forever and doing releases constantly. I personally have yet to see where SuSE has fallen down on that. If anything, it's a bit annoying that SuSE tends to lag other releases (2.2 kernel, glibc2, glibc2.1). Other stuff is apparently also a few releases behind (Gimp, for example). On the other hand, they've led the pack with XFree86 and ISDN (not that that's very useful to me or popular in the US, but I gather it's a big plus in Europe).
However, as far as stability, I've found it excellent. And there's a lot to be said for having the kitchen sink on the CD. Downloading and building stuff for the sake of downloading and building gets really tedious.
"Telnet is so slow..." -- what precisely is slow about it? Is it slow making the initial connection, slow at receiving output from the remote, or what? Is it inbound or outbound that's slow? Are you using a dialup or ethernet? What about your nameserver configuration?
I disagree about the manual; I think it's one of the best ones around. Certainly it's obvious that it's translated from German, with the occasional mistakes, German words or phrases, European phone numbers, and so forth, but everything I've read has been crystal clear, and the manual itself is exceptionally well produced (in terms of quality of paper, binding, and printing). To take one example, the description of how to set up PPP is actually clear and comprehensive.
All that aside (obviously I like SuSE), it's not for everyone. If you prefer Debian, that's what you should use. That's one of the big joys of Linux, choice! There's room for a lot of distributions, and it's nice to see Red Hat and SuSE going after each other (in the market) so aggressively.
Although I do wish they'd work with Linus to get their kernel additions (mostly drivers) back into the mainline. I hate trying to maintain two vendor branches in my CVS repository, and merging my own hacks back into both...
If it's your own program, there's no issue here. You can apply whatever licenses you choose: GPL, GPL with exceptions, BSD, proprietary... Note that I said "licenses". You, as copyright holder, can apply different licenses to different forms of distribution. You could GPL it for general use, and sell a version under a proprietary license with bundled support. You could give your company special exemption from the existing license rules, if that's what your intent is.
I'd be very surprised if Linux ran out of disk space on you and left you with an irrecoverable machine (requiring a disk wipe to recover). I won't say that it's impossible, but I would be extremely surprised. Remember, I'm not talking about losing the work that didn't fit, but a major system crash that essentially required an installation from scratch (and like someone else reported, this trashed his NT4 partition on the same drive).
This stuff about drivers overwriting memory that they shouldn't is exactly what the OS should handle *before* the attempted write. The driver should take what amounts to a seg fault if it tries to do that, and the Linux kernel can recover from a seg fault (it's when the recovery routine fails, or the fault occurs in certain critical code, that Linux panics).
The irony here is that one of the early knocks against Linux was that it's a monolithic kernel, while NT was supposed to be microkernel-based. In a real microkernel architecture, filesystems and drivers are usually implemented in user space; the kernel only manages processes, VM (at most) and the like. While Linux doesn't really fit the normal definition of a microkernel, since its filesystems and drivers run in ring 0 (and because Linus and Andy Tanenbaum say so!), it's interesting that almost everything except the boot device and filesystem can be modular (and even those can effectively be modules with an initrd), and that Linux can recover from many device and filesystem faults. In contrast, NT, where everything is supposed to be pluggable, seems prone to all sorts of device-oriented weirdness.
So how does the group and user stuff weigh down Unix? Because file and directory operations (not I/O itself) check permissions?
This sort of stuff is very useful even on a single-user machine, actually; it allows partitioning of privileges according to need. For example, there's no good reason for normal users to run as root routinely; even very experienced Unix (in the generic sense) use root privilege as little as possible. Why? It's too easy to screw something up, or for a rogue program to nail you.
You have a home network? Do you want your kids to have different privileges on the system from you, or to not have access to certain files? That's another way that multi-user is useful at home.
So what's so bad about an architecture that hasn't undergone serious revision (even if that were true, which it isn't)? One could argue (as I do) that that attests to its fundamental strength. It has certainly evolved over time: witness networking, VM, real time capability, and the many other things that have changed without doing fundamental violence to the underlying system. That strikes me as a system that was fundamentally rearchitected.
"...with Windows 2000, stability appears to be a non-issue." Huh? Windows 2000 isn't even GA yet; there is to be at least one more public beta, so it's hard to make any claims about its stability. What I've seen thus far is less than encouraging. A co-worker's machine (2000B2) crashed hard when his disk filled up and he tried to add more swap space (or something of the sort -- a relatively routine kind of thing, whatever it was). It took multiple OS reloads, none of which worked (apparently some registry issue, although no one ever figured it out), and eventually he had to wipe his disk and reinstall from scratch.
GUI-based OS: that's fine for desktops (until something goes wrong with the GUI, or if you don't like it). For servers, it's insane to be wasting resources making the console look pretty.
Third party proprietary drivers: there are several issues here. One is that that's one of the best arguments in favor of open sourcing the kernel, including everything running with elevated processor privilege (and a tacit admission that open source software IS more reliable, at least in that context). Also, a buggy driver doesn't necessarily crash the entire system; the Linux kernel is able to take an oops in a driver but keep running in many cases. This may also finally force a move toward lower-privilege device drivers.
"How long it will take Unix to catch up I don't know." I don't either, but maybe it won't be quite so long as you think.
As for IPC, certainly the open source community hasn't coalesced around a single standard as the MS community has (although the latter "standard" seems to change radically from release to release -- DDE gave way to OLE which gave way to COM/ActiveX which who knows what that will become). Gnome and KDE represent two approaches, and there seems to be a move toward rapprochement between the two camps.
XF86 crashes on you? I've had very few crashes indeed, and I can't remember any within the past year. Certainly I can't speak for you, but I'm not aware of other reports of flakiness. Perhaps you're using a poorly supported card?
"If Linux succeeds, perhaps in 20 years we'll see an uprising against Linux vendors." Linus is well aware of this, which is why he's trying to keep a handle on kernel bloat and cruft. Linux isn't the be-all and end-all of operating systems, and isn't intended to be. It's intended to be a better alternative (for suitable definitions of "better" for various purposes) to what's out there.
Expires on 20010630? Well, isn't THAT special!!??
on
DIVX is dead
·
· Score: 1
I found these statements very remarkable:
1) That all existing discs can be viewed only for the next two years, and (even more so)
2) That "upgraded" discs ALSO have that limit!
So it sounds like there's an "end of the world" switch built into DIVX, or else there's a fair bit of remote reprogramming that can be done on those things. Not Very Nice.
Here's what's wrong with "wide links = no" in this context: the test platform was intended to be a dedicated web server, not a general purpose timesharing system, so it should be configured to not permit random logins.
Yes, that setting has its uses, but not in this context.
Beyond that, I agree (very much!) that it's important for Linux to perform well on high end hardware, but in this particular case it wasn't really the issue. The particular RAID controller they chose was well supported on NT, but poorly supported at the time on Linux (although there was a newer driver). There are other RAID controllers available with equivalent capabilities that perform well on Linux, and furthermore the one in question is under development.
By default, the Linux kernel can't use more than 1 GB of RAM, but that wasn't an issue in the test system (it had 2 GB, but NT was also limited to 1 GB in the test). So that was neither here nor there. Also, as they pointed out, NT is limited to 4 GB, and Linux can be recompiled to use 2 GB. Considering that the Xeon can address 64 GB, both have a ways to go.
I would argue that IP rights exist ultimately as an agreement between parties. As long as I have a right to NOT produce or communicate the ideas I have come up with, as long as I have some right to privacy, then I can choose when and to whom I speak to -- if I speak my mind to one person, that does not require me to speak to all! If I offer to a second party, "I will tell you of the IP I have thought of, if and only if you give your word you will not repeat or reproduce it to any one else without my explicit permission," that is the creation of IP rights in essence. The second party may choose to refuse to accept the conditions of the transaction, but should the second part accept, then they are bound ethically to respect the my IP rights as I have claimed them. If the second party breaks their word, then that is reprehensible -- if transaction was made as a legal contract, enforceable.
That's a straightforward matter of contract; the notion of intellectual property doesn't enter here.
The summary of this transaction is:
You offer me valuable consideration (information that only you possess) in exchange for valuable consideration (my not making the information available to some other party without your consent).
I accept the terms offered.
You execute your side of the contract by giving me the information.
I violate the terms of the contract by giving the information to someone without your consent.
This is a straightforward contract violation, for which you can sue me. The amount of damages awarded would be whatever a judge or jury deems appropriate compensation for your loss of value.
The notion of implied contract with sale is interesting (and could in principle be enforced by requiring explicit agreement, although that might be unworkable -- an example of choosing between different "licensing" alternatives that offer different costs to the producer). However, that isn't the same thing as a copyright; under that system, if I legally hear a piece of music that I have not specifically agreed not to "copy", I can do whatever I like with it.
This kind of contract is more like a nondisclosure agreement, which is relatively noncontroversial. The problem is that the default rules governing information give the original producer rights that have nothing to do with contract between consenting individuals.
Note that nothing here is suggesting that you may not keep information secret -- it's just that you shouldn't have a default claim on it if you do choose to release it.
Why all this single-app threading? Multi-process!
on
Linux 2.3.0
·
· Score: 1
Threads aren't the only way of getting better SMP performance; full-fledged processes work just as well in most cases (except where latency really is the overriding issue), and it's an easier programming model.
For that matter, a machine running multiple jobs simultaneously will take advantage of multiple CPU's. Unfortunately, all the benchmarks (even server benchmarks) seem oriented toward single-tasking desktop platforms.
Slashdot Mirror
You're STILL entirely missing my point!
It's surely true that the vast majority of parents in the US wouldn't want their children to see what you described. Whether that would be true in all cultures is open to question. What's a web site provider in a culture that considers something like that normal to do?
That notwithstanding, you're also assuming that all web site providers are going to rate their sites in good faith and in accordance with your morals. Neither of these is going to be consistently true. There will be sites that intentionally misrate themselves, and there will be others that will be rated according to very different standards. The ratings on those sites will not even be useful as a starting point. Indeed, they will be worse, because they'll lull you into a false sense of security.
Part of my posting was explaining the DIFFERENCE between movie ratings and web site ratings. Movie ratings work in their own somewhat perverted way for several reasons:
1) There are relatively few movies made, by relatively few organizations, that achieve wide distribution.
2) There is a central ratings organization (in the US) that applies the ratings. This is possible because of the relatively few movies made, but also because relatively few places exhibit movies and they all have much the same policies.
3) The central ratings organization only rates movies for the US (other countries have their own boards), which is fairly homogeneous culturally (compared to the world at large).
4) Movies are static content (they don't change constantly -- if you see Titanic today, it's the same movie you saw a year ago).
NONE of these conditions are true for the web! There are probably on the order of six orders of magnitude more web sites than there are movies made in any one year. That's a factor of a million. It is not possible for one organization to "screen" all of these sites, so ratings will have to be applied by the site operators. Many of these sites are not located in the United States, so ratings will be applied according to the standards of the operators of those sites. And finally, sites change constantly.
You then have to take into account that there's no organized web site industry that can enforce this kind of cartel, as each web site operator is independent of every other operator. There are web site providers who are malicious -- who WANT to expose your children to pornography. What's going to happen when one of these sites is rated G? Are you going to relax your guard, since the G rating means that it's good for your children, or will you still have to prescreen it against these malicious operators?
And before you say that these folks are breaking ze law and must be put in prison, you first have to find them. Tracking these folks down is hard, especially when they live and provide service overseas. And speaking of overseas, what about people overseas who simply have different values from you and think that full frontal nudity is a normal part of life and that there's no reason children should not be exposed to it?
My point is that this "enforced ratings system worldwide" simply cannot work, so there's no use talking about any potential benefits or uses of it. It doesn't matter what you think the potential benefits may or may not be; since the system cannot possibly work as you intend, these benefits don't exist.
The ONLY way that filtering can work is on the client side. Whether you do it personally (which has the highest accuracy level) or you use software produced by someone you trust, or you rely on voluntary ratings (which can be subverted unless there's some real authentication behind it, which still requires what amounts to client-side verification), THIS MUST HAPPEN AT YOUR END! You CANNOT rely on the server to give you ANY useful information at all about appropriateness of the content. Period.
Now do you understand my point? What you're asking for cannot exist, so the whole basis for your argument is flawed.
You haven't answered any of my objections, merely reiterated your original message. There's absolutely no way that "mandatory ratings" can be enforced; content providers can always disagree with your interpretation (intentionally or otherwise); and how does this help parents who have different ideas about what matter?
Remember that this is very different from the movie industry, where everyone's cooperating (many would say "colluding" or "conspiring") to make the rating system "work" after a fashion, or television, which is basically controlled by a few large corporations (and even with cable there are only a relatively few discrete channels). On the internet, there are MILLIONS of content providers who really don't care about your personal values, and some who would probably like to actively subvert them. This kind of thing simply won't work on the net.
If you really want to CONTROL what your children see on the net, simply arrange to be there at all times with them. There's no other way to do it. If that doesn't work, learn to live with the fact that the world is imperfect and find a better way to deal with it. There's no more gentle way of putting it.
It's bad enough that movies are de facto mandated ratings (theaters won't show unrated or NC-17 movies). As for video games and comics, are they rated (in the US, at least)? I wasn't aware of that. If people WANT to rate themselves, that's fine. If not, it's another matter.
One obvious problem with mandatory ratings of web sites is: who's going to do the rating? There are so many web sites and they change on such a continuous basis that there's no possible way that a centralized rating authority can assign ratings to every web page (or really every separately addressable object) in existence. That means that content providers have to supply their own ratings. There's obviously a conflict of interest here. The Australian law seems to try to get around that by requiring that the ratings coincide with what the central censorship bureau would assign the site. That makes it difficult, to say the least, for people to safely rate their own sites. This issue has been discussed in a lot more depth elsewhere, and I don't feel like a lot of typing.
As for this issue of parents vis a vis children, if you want your children to be that coddled, then you can bloody well find the time to research every single thing your child wants to buy/see/use. If you don't want to, you can buy filtering software to do the job for you. If you complain that it doesn't work well enough, then write better software yourself to do it. But don't ask the person running the erotica site to admit that his material is morally wrong and to do your work for you. If nothing else, it won't work. Mandating that every web site be correctly rated is impossible. Don't pretend that it will work -- it won't, and your child will be exposed to all the "garbage" (distinctly a value judgment) that he'd be exposed to anyway. If you think that "mandatory" ratings will relieve you of any effort, you're fooling yourself, I'm afraid.
Besides which, ratings are a very limited way of measuring something. I presume you're thinking about ratings for erotic and violent contact. But what about an Islamic parent who doesn't want his son to see pictures of women with their legs uncovered? Or a Jewish parent who doesn't want her children to get all caught up in Christmas? How does any rating system "protect" them?
Isn't it better to teach your children how to properly handle material that you or they consider undesirable than to pretend that you can magically stop it?
I've been hacking on a driver for the Epson Stylus Photo EX for a while. This is actually Michael Sweet's driver for the Gimp print plugin, ironically enough.
:-) ) could tune the output for different papers.
It turns out that Epson has (at least) two levels of their developer documentation, beyond the standard documentation for their printer language (ESCP/2). The "Level 1" documentation on their developer web site (http://www.ercipd.com/isv/edr_docs.htm) gives fairly detailed documentation on the printer, but there's at least one mroe level of documentation that's available only under NDA, which (obviously) I don't have. So I've been able to get very good print quality, but not the level of quality the printer is ultimately capable of. I've found another Ghostscript driver around that apparently does know how to achieve the higher print quality (looking carefully at the printer output it's obvious -- with an 8x loupe and simply looking at the micro-banding, it's clearly doing something different), but I haven't had time to study the code yet (my driver seems to produce better fidelity).
BTW, if anyone would like to hack on my modifications to Michael's Gimp plugin, it's available at http://www.tiac.net/users/rlk/print.tar.gz. I'd like to find more people to test this and make suggestions; it's become a rather expensive hack (in terms of paper, and particularly ink).
Whatever my thoughts about the multiple levels of documentation, though, I don't particularly blame Epson for putting the smarts in a software driver. Dithering algorithms and such for 6 colors (CMYK + light C + light M) seem to become very complex very quickly, and having this stuff be under software control makes it much easier to experiment with changes. For example, my driver achieves higher color fidelity (visually compared to my monitor) than the driver Epson bundles. It actually makes the printer a MORE open device than if the firmware were embedded. I wish Epson would put the entire programming information on its web site, not just parts of it, but at least I (or others who know more about color and dithering than I do) have a chance of reverse engineering the format and improving the output quality beyond what even Epson achieves. A true PostScript printer certainly has a very open interface, but a closed implementation. As much CPU as this consumes, I think that for printers such as this, having an open implementation is very valuable.
Examples of improvements that could be made:
1) More user-level controls (I've added gamma, contrast, and color level controls to the existing set). This enables fine tuning for a particular image or monitor.
2) Different kinds of paper. Images print very differently on different papers. Photo-quality glossy paper is different from "photo-quality" coated paper, which in turn is much, much different from standard ink-jet paper. This is mostly due to how much ink the paper absorbs. People who actually know what they're doing here (as opposed to dabblers such as myself who like to shoot off our mouths
3) Different inks. I bought a third party cartridge because it was the only one I could find at CompUSA. Its color balance was somewhat different. Conceivably a third party might devise archival inks (that don't fade rapidly) that have different colors and densities. With a software driver the printer really does become open; if it's in ROM on the printer, there's no adjustment for it.
4) Improved dithering algorithms. Someone might devise better dither patterns to get smoother output.
Red Hat did exactly what you claim Debian did (I don't actually know what Debian did) -- they allow use of Red Hat in the name (e. g. "derived from Red Hat Linux" or whatnot), as long as it's clear that it's not the official package. I don't have the URL to the letter that they sent out about it, but it got heavily publicized on Slashdot and Linux Today, among others.
Red Hat acted honorably here. They want to use the trademark to make it clear to customers what they're getting, not to screw anybody (except people who wanted to pretend that they were selling the official package with support, as opposed to a cloned CD).
It simply means that an institution of higher learning has chosen to grant somebody a high honor for whatever reason of their choice. Some places (e. g. MIT) don't grant them at all; others (such as the little red brick community college up the creek, sometimes known as Harvard) do.
I don't believe that it entitles him to use "Dr." as a title. So in some ways it means more, and in more ways it means less, than a true PhD. But it certainly indicates that a (in this case prestigious) institution of higher learning has seen fit to bestow a major honor on him. To reject it (as a few people suggested) would be a major insult, I should think.
Linus has certainly done a lot of good for a lot of people. See, for example, the project in Mexico placing computers running Linux in classrooms -- Microsoft software is simply too expensive to permit that. One can argue about how much code Linus has written vs. the FSF vs. Berkeley vs. the other kernel hackers, but there's little question in my mind that he did considerably more than just be in the right place at the right time and go along for the ride.
Congratulations, Linus!
One will note that Linus's message announcing 2.3.18 merely indicated that he would be away for a few weeks and not reading email. It implied nothing more than a well-deserved vacation after having feature frozen 2.4. He made absolutely no mention of the honorary doctorate.
Somehow, this doesn't strike me as the mark of a pompous ass.
My father was a chemical engineer. He did have a PE certification in a couple of states, but it very seldom mattered during his employment. After he retired and went into consulting, though, it did matter for certain jobs, for exactly the reasons specified -- a lot of his work involved investigations after accidents and the like, and he frequently had to write reports that would have legal standing in the event that there were any lawsuits over, say, denied insurance claims.
My father was only licensed to practice in two states. As it happens, there are only a few states that matter from the standpoint of the particular subspecialty of chemical engineering that he practiced.
Chemical engineering is a very traditional field where pieces of paper mean a lot and (at least 15 years ago) everyone wore suits to work every day. Even so, it sounded like his PE really only mattered very much when he was consulting.
The PE examination requires substantial knowledge in many areas -- electrical, chemical, civil, and such. For the vast majority of people in the software (and even hardware) field who style themselves (or are styled by their companies) "engineer", this has no relevance whatsoever. I, for example, mostly do OS and other low level work. I'd be surprised if there are more than a handful of people in the building who even know what a PE is, much less what one does.
Most software vendors seem to subscribe to the "we don't accept liability" school of thought. NT's EULA, which apparently states that NT is not certified for use in safety-critical situations, is a case in point. If they get sued over somebody's death because of an OS failure, they're simply going to point to the EULA and state that they made it quite clear that this is not qualified for the purpose. Of course, any PE at the nuclear power plant who signed off on using such a product would probably be in trouble, but any software vendor that takes that kind of approach probably doesn't need a PE on staff.
EE's who design consumer electronics probably don't need a PE at all. Most contractors probably don't, either. However, if you want contract jobs in situations where you might be called to testify in court (such as investigations of problems after the fact), or where you have to certify that something was done according to best accepted practice, it's probably essential. So companies that write custom software to run nuclear power plants probably do need some people with PE certification.
I wrote a letter of nomination for Alan before I read this discussion. Enclosed forthwith.
I hereby nominate Alan Cox for the 1999 Free Software Award
nomination.
Alan Cox is perhaps Linus Torvalds's chief lieutenant. I think Alan
should be recognized, however, for a very specific task that he has
willingly taken on: maintenance of old stable releases of the Linux
kernel. This is not a glamorous task, but is one that is essential to
counter the impression that free software lacks support.
Alan has taken it upon himself to maintain the 2.0 and 2.2 Linux
kernels while others forge ahead with the more exciting work of
designing new features for the development kernels. While the
stability of Linux 2.0 is legendary, bugs are occasionally discovered,
new devices become available, and security holes are found. In order
to be accepted for mission critical applications, old versions of
software must be supported, with critical items such as these fixed.
Responsible software vendors have always provided such support, and
Alan's work has made it possible for free software to be held to such
exacting standards.
In addition to this key responsibility, Alan is also a major
contributor to the development kernel, and particularly concerns
himself with low level kernel functionality and devices, which are
notoriously complex yet essential pieces of functionality.
Finally, Alan ceaselessly promotes free software in all forums in
which he participates.
For these reasons, I believe that Alan Cox would be a deserving
recipient of the 1999 Free Software Award.
Actually, even in the heart of the tornado belt, any one location is struck on average once every 2000 years. The repeat frequency for large hail or destructive wind is, of course, much more frequent. However, the important point is that there's no location within that zone that's particularly more or less prone to devastating effects.
Residents of barrier islands and the immediate coast in hurricane country (the entire east coast, in other words) are in quite a different situation. Significant storm surge isn't as rare as that, and what's more, there are well-understood reasons why those locations are much more vulnerable than even a few miles inland. Building somewhere that any competent meteorologist or structural engineer can tell you is a sitting duck for a fairly common kind of storm is asking for trouble.
Before people talk about Homestead and Andrew, let me point out that Homestead was really at ground zero for a particularly devastating storm; that quite a few well-constructed buildings did survive; and that there was evidence of a previously unknown phenomenon (extreme wind gusts resembling microbursts in the eyewall) that is hard to defend against by any means, and that really is very localized and random.
Storm surge affects a much greater frontage of shore, and is far more devastating than even a hurricane's rain and wind. There's also a simple defense against it -- just don't build anything close to certain kinds of shoreline that you're not willing to have destroyed every so often. Nothing will really stop a storm surge other than distance. Winter storms also cause destructive surges from Cape Hatteras north, so there's no excuse for excessive development too close to the northern shoreline (there's also less of it, because it's less desirable, and because winter storms are much more common).
It's not only hurricanes that are an issue -- anyone who builds a house in a known floodplain, or close to a volcano that might be active, or in an obvious mudslide channel is also being pretty silly, when it's not hard to choose a much safer location nearby.
Meanwhile the storm is still tracking to the left of the forecasts. Check out ne.weather and sci.geo.meteorology for a lot of discussion on it. It's starting to turn (seemingly fairly decisively finally), but it's going to be an awfully close shave at best.
Probably very few people here (I number among the "many") have ever seen a major hurricane make a close pass. Here in the northeast, the only hurricanes I've seen were David (1979), Gloria (1985), and Bob (1991). Each of these were fairly weak by the time they reached us. If Floyd does come anywhere close to the location of the webcam, there will be a lot of opportunities to see what this kind of storm can really do.
The truth of the matter is that a major hurricane can hit anywhere along the east coast and the Gulf of Mexico, and in the northeast, if a major hurricane does hit, it can carry strong winds a considerable distance inland (because such a storm is invariably moving very fast -- 40 MPH forward speed or greater -- and simply doesn't have time to weaken, and furthermore the right hand side of such a storm has its winds enhanced by the forward motion of the storm). Hazel in 1954 went inland at South Carolina and screamed north at 50 MPH or so, so it had been inland for 500 miles or thereabouts by the time it reached New York's latitude. Nonetheless, it spawned a wind gust of over 100 MPH at the official observing station (probably Central Park), which was well east of the center. The 1938 hurricane was cat 3 (still a major hurricane) when it reached the coast, and Blue Hill had a peak wind (1 minute, which is not a gust) of 183 MPH in that storm!
Anyway, this storm might nail (to some degree) the entire east coast. If the storm does hit hard, and if the webcam survives, it might help warn other people what to really expect from a major storm. It's weakening a bit (down to 125 knots from 135 last night), but not enough to matter. The one piece of good news for Florida is that if it is a coast hugger the coast will be hit with the less severe west side of the storm. However, it's an absolutely gigantic system, so even a moderately near miss will cause problems. It's not a tiny system like Andrew was.
I'm also on the "have a core of employees for core product" side, for reasons that people have elucidated. While I agree that even employees should not assume absolute job security, I don't believe that most companies are going to take that short term of a horizon toward their employees. And in some ways, if things turn bad an employee is in a better position to up and leave than a contractor. Leaving in the middle of a contract doesn't strike me as a terribly wonderful thing to do, whereas employees are always free to leave whenever they please.
Furthermore, this attitude "everyone's short term anyway, so I might as well be a contractor, rake in the $$$, and keep jumping aroune" is exactly wrong for a company trying to build up a long term base of experience (which should be any company trying to build a product, as opposed to a company that's merely a body shop). At my previous company, we scrutinized very carefully anybody who seemed to jump around a lot. Our CEO firmly believed in employees rather than contractors for all but very clearly defined, very specific projects, and we all agreed with him on that issue.
I can think of one particular exception at an earlier job, though. This was an individual who wanted a contract specifically for benefits reasons -- he had some medical condition that was well covered by his insurance, and he didn't want to risk having any coverage problems being on employee health benefits and preferred the extra cash to useless health insurance. However, it was understood that he was in all other respects an employee. That worked out, but the important thing is that he had an employee attitude, and wished to contract for unique reasons related to benefits.
Under no circumstances would I hire a contractor as a lead system administrator. I'm certain that there are plenty of them who do conscientious work, and I'm well aware that there are disgruntled employees, I would much prefer to have a system administrator whose incentives and mindset are structured toward a long-term relationship with the company. Certainly I would permit hiring a contractor for specific tasks, such as assisting in a major network expansion or reconfiguration, but the person who's in charge of systems and security is even more important to keep around and suitably gruntled than, say, a VP of engineering.
Anybody who tries to compare any PC to a Starfire is obviously going to look silly -- if you're running a workload that the Starfire can run effectively (nicely parallel). But then again, throw together a Beowulf cluster with enough cheap PC's and run something that's reasonably close to embarrassingly parallel, and you're right up there in the list.
Somehow, it seems awfully cynical to call anything that's in the 3-digit mips and Mflops range a "puny toy". Any of these things has the raw CPU to blow a Cray-1 out of the water (of course, it doesn't have the memory bandwidth, but that's another matter). It's also a very large fraction of what we could do on a Connection Machine (CM-2) 10 years ago. Sure, I'm well aware that I can spend less on a processor than on the latest Star Wars toy for my nephew, and likewise that processor technology's advancing at an incredible rate, but these chips would have been called supercomputers not too many years ago.
And then I shudder to think how these things actually slow down as much as they do running Windows, and I can't imagine what must be going on that mere UI display (without any speech recognition) can bog one of these things down. Oh well, I guess I can just be smug about the fact that I can't even *run* Windows any more, since my last mobo upgrade...
Computernerd.com isn't pretending that their Celerons are PII's, so what fraud is involved? As far as 300/450A's vs. 466's, if you really can clock a 466 to 700 MHz, then yes, it will beat the 300/450, but unless you can do that, the 450 will win anything that goes significantly out of cache (just about everything except a synthetic computation kernel). Why? Because Celerons are overclocked by boosting the FSB frequency from 66 MHz to something else (100 for the 300/450, for example). This reduces the memory bottleneck substantially.
Certainly it's a firing offense anywhere (and perhaps grounds for a defamation lawsuit). The question is whether Microsoft actually wants to go after this individual or not.
Notice that:
1) Thus far they have been unable to identify the employee in question (which if their gateway is even halfway decent shouldn't take more than a few hours at most).
2) Rob Bennett (director of marketing, MSN) didn't exactly condemn whoever did it "somebody who got a little over-passionate but went about it the wrong way." Sounds like praising with faint damnation to me, especially in a culture that rewards taking risks.
3) Not only does Bennett not condemn the unnamed employee, he concurs with the statements so made.
It probably (IMHO) was not officially sanctioned, but I suspect that there was no need to actually ask someone to do so or formally approve it.
There needs to be a balance between fixing bugs forever and doing releases constantly. I personally have yet to see where SuSE has fallen down on that. If anything, it's a bit annoying that SuSE tends to lag other releases (2.2 kernel, glibc2, glibc2.1). Other stuff is apparently also a few releases behind (Gimp, for example). On the other hand, they've led the pack with XFree86 and ISDN (not that that's very useful to me or popular in the US, but I gather it's a big plus in Europe).
However, as far as stability, I've found it excellent. And there's a lot to be said for having the kitchen sink on the CD. Downloading and building stuff for the sake of downloading and building gets really tedious.
"Telnet is so slow..." -- what precisely is slow about it? Is it slow making the initial connection, slow at receiving output from the remote, or what? Is it inbound or outbound that's slow? Are you using a dialup or ethernet? What about your nameserver configuration?
I disagree about the manual; I think it's one of the best ones around. Certainly it's obvious that it's translated from German, with the occasional mistakes, German words or phrases, European phone numbers, and so forth, but everything I've read has been crystal clear, and the manual itself is exceptionally well produced (in terms of quality of paper, binding, and printing). To take one example, the description of how to set up PPP is actually clear and comprehensive.
All that aside (obviously I like SuSE), it's not for everyone. If you prefer Debian, that's what you should use. That's one of the big joys of Linux, choice! There's room for a lot of distributions, and it's nice to see Red Hat and SuSE going after each other (in the market) so aggressively.
Although I do wish they'd work with Linus to get their kernel additions (mostly drivers) back into the mainline. I hate trying to maintain two vendor branches in my CVS repository, and merging my own hacks back into both...
If it's your own program, there's no issue here. You can apply whatever licenses you choose: GPL, GPL with exceptions, BSD, proprietary... Note that I said "licenses". You, as copyright holder, can apply different licenses to different forms of distribution. You could GPL it for general use, and sell a version under a proprietary license with bundled support. You could give your company special exemption from the existing license rules, if that's what your intent is.
One particular NT 3.51 configuration achieved a C3 rating (certainly not B2) when it was not connected to any network.
I'd be very surprised if Linux ran out of disk space on you and left you with an irrecoverable machine (requiring a disk wipe to recover). I won't say that it's impossible, but I would be extremely surprised. Remember, I'm not talking about losing the work that didn't fit, but a major system crash that essentially required an installation from scratch (and like someone else reported, this trashed his NT4 partition on the same drive).
This stuff about drivers overwriting memory that they shouldn't is exactly what the OS should handle *before* the attempted write. The driver should take what amounts to a seg fault if it tries to do that, and the Linux kernel can recover from a seg fault (it's when the recovery routine fails, or the fault occurs in certain critical code, that Linux panics).
The irony here is that one of the early knocks against Linux was that it's a monolithic kernel, while NT was supposed to be microkernel-based. In a real microkernel architecture, filesystems and drivers are usually implemented in user space; the kernel only manages processes, VM (at most) and the like. While Linux doesn't really fit the normal definition of a microkernel, since its filesystems and drivers run in ring 0 (and because Linus and Andy Tanenbaum say so!), it's interesting that almost everything except the boot device and filesystem can be modular (and even those can effectively be modules with an initrd), and that Linux can recover from many device and filesystem faults. In contrast, NT, where everything is supposed to be pluggable, seems prone to all sorts of device-oriented weirdness.
So how does the group and user stuff weigh down Unix? Because file and directory operations (not I/O itself) check permissions?
This sort of stuff is very useful even on a single-user machine, actually; it allows partitioning of privileges according to need. For example, there's no good reason for normal users to run as root routinely; even very experienced Unix (in the generic sense) use root privilege as little as possible. Why? It's too easy to screw something up, or for a rogue program to nail you.
You have a home network? Do you want your kids to have different privileges on the system from you, or to not have access to certain files? That's another way that multi-user is useful at home.
So what's so bad about an architecture that hasn't undergone serious revision (even if that were true, which it isn't)? One could argue (as I do) that that attests to its fundamental strength. It has certainly evolved over time: witness networking, VM, real time capability, and the many other things that have changed without doing fundamental violence to the underlying system. That strikes me as a system that was fundamentally rearchitected.
"...with Windows 2000, stability appears to be a non-issue." Huh? Windows 2000 isn't even GA yet; there is to be at least one more public beta, so it's hard to make any claims about its stability. What I've seen thus far is less than encouraging. A co-worker's machine (2000B2) crashed hard when his disk filled up and he tried to add more swap space (or something of the sort -- a relatively routine kind of thing, whatever it was). It took multiple OS reloads, none of which worked (apparently some registry issue, although no one ever figured it out), and eventually he had to wipe his disk and reinstall from scratch.
GUI-based OS: that's fine for desktops (until something goes wrong with the GUI, or if you don't like it). For servers, it's insane to be wasting resources making the console look pretty.
Third party proprietary drivers: there are several issues here. One is that that's one of the best arguments in favor of open sourcing the kernel, including everything running with elevated processor privilege (and a tacit admission that open source software IS more reliable, at least in that context). Also, a buggy driver doesn't necessarily crash the entire system; the Linux kernel is able to take an oops in a driver but keep running in many cases. This may also finally force a move toward lower-privilege device drivers.
"How long it will take Unix to catch up I don't know." I don't either, but maybe it won't be quite so long as you think.
As for IPC, certainly the open source community hasn't coalesced around a single standard as the MS community has (although the latter "standard" seems to change radically from release to release -- DDE gave way to OLE which gave way to COM/ActiveX which who knows what that will become). Gnome and KDE represent two approaches, and there seems to be a move toward rapprochement between the two camps.
XF86 crashes on you? I've had very few crashes indeed, and I can't remember any within the past year. Certainly I can't speak for you, but I'm not aware of other reports of flakiness. Perhaps you're using a poorly supported card?
"If Linux succeeds, perhaps in 20 years we'll see an uprising against Linux vendors." Linus is well aware of this, which is why he's trying to keep a handle on kernel bloat and cruft. Linux isn't the be-all and end-all of operating systems, and isn't intended to be. It's intended to be a better alternative (for suitable definitions of "better" for various purposes) to what's out there.
I found these statements very remarkable:
1) That all existing discs can be viewed only for the next two years, and (even more so)
2) That "upgraded" discs ALSO have that limit!
So it sounds like there's an "end of the world" switch built into DIVX, or else there's a fair bit of remote reprogramming that can be done on those things. Not Very Nice.
Here's what's wrong with "wide links = no" in this context: the test platform was intended to be a dedicated web server, not a general purpose timesharing system, so it should be configured to not permit random logins.
Yes, that setting has its uses, but not in this context.
Beyond that, I agree (very much!) that it's important for Linux to perform well on high end hardware, but in this particular case it wasn't really the issue. The particular RAID controller they chose was well supported on NT, but poorly supported at the time on Linux (although there was a newer driver). There are other RAID controllers available with equivalent capabilities that perform well on Linux, and furthermore the one in question is under development.
By default, the Linux kernel can't use more than 1 GB of RAM, but that wasn't an issue in the test system (it had 2 GB, but NT was also limited to 1 GB in the test). So that was neither here nor there. Also, as they pointed out, NT is limited to 4 GB, and Linux can be recompiled to use 2 GB. Considering that the Xeon can address 64 GB, both have a ways to go.
That's a straightforward matter of contract; the notion of intellectual property doesn't enter here.
The summary of this transaction is:
This is a straightforward contract violation, for which you can sue me. The amount of damages awarded would be whatever a judge or jury deems appropriate compensation for your loss of value.
The notion of implied contract with sale is interesting (and could in principle be enforced by requiring explicit agreement, although that might be unworkable -- an example of choosing between different "licensing" alternatives that offer different costs to the producer). However, that isn't the same thing as a copyright; under that system, if I legally hear a piece of music that I have not specifically agreed not to "copy", I can do whatever I like with it.
This kind of contract is more like a nondisclosure agreement, which is relatively noncontroversial. The problem is that the default rules governing information give the original producer rights that have nothing to do with contract between consenting individuals.
Note that nothing here is suggesting that you may not keep information secret -- it's just that you shouldn't have a default claim on it if you do choose to release it.
Threads aren't the only way of getting better SMP performance; full-fledged processes work just as well in most cases (except where latency really is the overriding issue), and it's an easier programming model.
For that matter, a machine running multiple jobs simultaneously will take advantage of multiple CPU's. Unfortunately, all the benchmarks (even server benchmarks) seem oriented toward single-tasking desktop platforms.
The job mix is the key!