MS's LinuxMyths says:
Linux does not provide support for the broad range of hardware in use today; Windows NT 4.0 currently supports over 39,000 systems and devices on the Hardware Compatibility List. Linux does not support important ease-of-use technologies such as Plug and Play, USB, and Power Management.
(Note that Windows NT 4.0 didn't support Plug and Play either. A curious instance of slamming Linux for missing something that NT was also missing.)
Ironically, with Windows 2000, Microsoft's offerings have gotten worse in some dimensions; the hardware compatibility list has shrunk roughly in half.
The application story for Windows 2000 is similarly weaker since Microsoft has traded off backwards compatibility in an attempt to improve stability. A significant fraction of Windows NT 4.0 applications don't run under Windows 2000 due to new restrictions imposed to improve stability, for example, restricting applications' ability to write freely into the registry (oops, now there was a security/stability hole!).
The judge's recommendation is a little more subtle. Six months and a day later, he suggests that your email isn't admissible if it is the sole piece of evidence. But if there is a continuing pattern of conversation or if there is evidence you are acting (not just talking) in some way as to substantiate your threat, the statute of limitations reverts to the time period currently proscribed by law.
It's a nice idea. It caused me to reflect on all the times I haven't posted on USENET or other places under my own name for precisely the reasons he's mentioned. Fear of my words coming back to haunt me (due to technical inaccuracy or being taken out of context, read by a future boss, etc.) has definitely led me to post less at times. Then again, maybe this is a good thing.;-) But I agree; lack of a time horizon for computer-mediated communications definitely leads to self-censorship, something we should be wary of.
Did IRIX really start as a fresh implementation as the chart describes? I thought it was based off the BSD releases, based on early data sheets I have on IRIX 3.x...
If you're interested in JavaSpaces and distributed.net-type computing, you might check out http://www.popularpower.com/. It's hard to tell from the webpages whether it uses JavaSpaces per se, but it clearly uses Java virtual machines as the sandbox for distributing parallel computations across thousands of nodes on the Internet, and they do have Windows and Linux clients.
The current workload appears to be a pure volunteer effort aimed at helping develop better flu vaccines, but once installed, obviously other (for-pay) algorithms implemented in Java could be run on the client, with Java's sandbox preventing you from virus-type problems. Interesting stuff.
The other issue is that RAM prices have stabilized after huge drops from late 95 to early 99 that far exceeded the pace of Moore's law. RAM manufacturers were losing their shirts and not willing to double down investing in huge new factories back then that would be reducing the shortage now.
The pendulum shortages/crashes of commodity devices like RAM are one reason why Intel got out of the business in the late 1970s.
--LP
Disclaimer: I've tracked RAM prices on pricewatch for several years but my analysis is more conjecture based on reading trade press than first-hand knowledge.
I found it ironically funny that, checking DNS records, the owner of the obvious domain name for MS's latest techno-marketing initiative, dot.net, is a guy from Sun.
Domain Name: dot.net Administrative Contact, Technical Contact, Zone Contact: Comay, David S (DC115) David.Comay@ENG.SUN.COM
This must be another example of Microsoft "leadership" and "innovation" at work...:)
I find it tremendously ironic that nobody has actually pointed out in 150 posts that the content of the article is, unfortunately false. It's vaporware guys. Did any other prospective developers actually go to the site to try to download such code? I did. There is no Game Construction Toolkit to download. Instead, their site says:
... The program supplies a WAP Client Toolkit, a Game Construction Toolkit, Application Programming Interfaces (APIs), documentation and sample source code for applications.... The Developers Program's toolkits will be available to registered developers from Autumn 2000. Before that, you can download our Developer's Guide and other resource documents, or read our selection of FAQ relating to the Mobile Entertainment Service.
Yawn. What a disappointment. Here I wanted to see how quickly I could port my 3D "tetris" game to a cell phone and I have to wait another four months. This is a good reminder of how poor the quality of Slashdot information and Slashdot community information is. Well, here's one member of the community trying to get the facts I've learned out. We'll see if you moderators cooperate.:-) (Is my criticism of Slashdot too insightful for you? Oh, sorry, I'm not supposed to mention the invisible moderators in a plea for points, either overt or reverse- sychologically, right?)
As a matter of fact, strict compatibility actually raises the value of all products, including those from Microsoft. Given that fact, it makes no sense for Microsoft to create an incompatible version of Kerberos.
Your first sentence is correct, your second is not. All products do benefit when they are compatibile and interoperable with one another. But Microsoft doesn't want to raise the value of all products. It only wants its own to benefit.
It makes perfect business sense for Microsoft to try to lower the value of competing products by preventing interoperability with its own. It's called lock-in, and it increases switching costs for users and barriers to entry for competitors. It's a strategy that makes perfect sense if you have a dominant (especially monopoly) position, and little or no sense if you don't have such a position. Harness network effects to exclusively benefit your product, what could be simpler?
The only time this doesn't pay off is if it sufficiently alienates customers or developers. So far, Microsoft has managed to hew a fine line where such alienation has not outweighed the benefits of its platform. It's up to knowledgeable people to point out the oft-hidden costs and risks of adopting Microsoft's technology approach.
Remember, in reality, most of Microsoft's succesful innovations have been *legal* innovations, beginning with their DOS contract and extending through various exclusive OEM agreements and their chiseling away at the Java contract and DOJ Consent Decree. Their trade-secret licensing of Kerberos and their attempt to license software on a renewal basis (first at universities) are just the latest examples of this. Just what you'd expect from a firm founded by the bright son of a lawyer.
...does not constitute contributory infringement if the product is widely used for legitimate, unobjectionable purposes...
This would seem to be the weakest argument of the Napster case; if the RIAA can show Napster is not widely used legitimately, do they win the case? There are legal Napster users out there (like myself, although I don't use Napster a lot), but I can't imagine there are too many of us.
Making available for public download is legally the same as publishing.
Hmm, while this rings true, I guess I'd have to look up a legal definition of publishing.
If I (a Napster user) never let anyone actually download, (a function supported by Napster that is not available in otherwise analogous FTP server software) then the premise of the statement (that the music on my hard disk is, in fact, available for public download) is false. Files on my hard disk do show up in Napster listings, but I haven't, as a matter of actual practice, had or let people download them.
It requires them to convince a judge that putting files on Napster constitutes publishing. I don't think they'll have any problems.
Perhaps. I suppose I would likewise view my point as picayune if either A) I'd never used Napster and relied on others to describe it to me, or B) I typically used it illegally. However, from the vantage point of a user attempting to take advantage of the software to move my (legal) music collection onto my hard disk, the distinction is highly relevant.
(I didn't find the DeCSS "legal excuse" that the software was for viewing DVDs on Linux rather than for cracking encryption and piracy very compelling myself, but then, that's partly because I wasn't part of the crowd trying to legally view DVDs on Linux.)
Napster said that they would block any user pirating MP3s on their system. They just needed a list. So this is what Metallica did.
Not exactly. You're overlooking something. Metallica gave them a list of people who showed up in a list as potentially offering the song for download. *Offering*, not *pirating*. There's a difference, a crucial one for a (gasp) legal user of Napster such as myself. (I have used Napster to get recordings on my computer that I have already purchased in cassette form.)
Offering songs for download is not a crime. Actually transferring them is. It's quite possible (and has been true for me in my admittedly limited Napster usage) that one can "list" a copyrighted song without it ever being transferred, particularly if one likes obscure bands or does not stay online with Napster very long.
Napster also provides chat facilities and the ability to detect that someone is downloading your file and the ability to boot them. These enable me, the purchaser of the product, to enforce the copyright provisions that I have always been responsible for as with previous technologies by checking to see whether the recipient is pirating rather than getting a legal (fair use) copy from me. Given the way the Napster software scans my hard disk and does not offer me an up-front choice of exactly which songs to list, this is a quite-likely possibility for someone with a mix of free (i.e. Grateful Dead) recordings mixed with a list of non-free ones who legitimately and legally wants to use Napster to share music while storing his/her own music on their computer.
To accurately and properly accuse 335,435 users of pirating (breaking copyright), Metallica would actually have to demonstrate that each one of them had not just offered a file, but transferred it to someone else illegally (in the process, verifying that said file was, in fact, a Metallica song for which they held the copyright.)
So I see this as one big PR move in an attempt to intimidate Napster and Napster users. Those 335,435 names (handles, actually) are nothing more than a huge list of "subversives" using the "wrong" software, not a list of 335,435 criminals. Innocent until proven guilty (in the US), right?
Now, the *real* problem with the Napster approach: But just because a legal loophole may knock this issue off the table, don't get arrogant. The labels can subvert Napster whenever they want by flooding Napster with falsely-named files, if they ever get enough clue and guts to do so (hiding their fingerprints to avoid backlash by pointing to those anarchist internet trolls as the culprits.) For example, instead of a real track, put a 30-second promo clip appended with 3-5 minutes of white noise. There are countermeasures to this, but ultimately the only real solution to limiting unjust corporate power is for citizens to either stop buying or organize politically and press for *real* "fair use" copyright provisions.
...making those MP3s available for public download (what you call trading) happens to be a copyright violation...
Close, but technically false. Making them available for download is not a crime. Actually transferring them is. It's quite possible (and has been true for me in my Napster usage) that one can "list" a copyrighted song without it ever being transferred, particularly if one likes obscure bands or does not stay online with Napster very long.
For Metallica to claim I broke their copyright requires them to show I transferred a file to a person who did not own a legit copy. To have a real legal case (AFAICT, IANAL), they would have to mount a sting for each person so accused: attempt to actually download a file from me (and confirm that the file really was "their" content).
--LP
RedHat ignores history of GNU contribution
on
Thus Spake Stallman
·
· Score: 2
I agree and would only point out one more thing.
Note how Red Hat ignores the contributions of GNU (the GPL, gcc, core UNIX tools) in their version "Linux History." For someone who watched and played with Linux in 1992-1993, this ignores the major way in which GNU tools enabled Linus and collaborators worldwide to develop Linux. I've talked to the marketing manager personally about this (as an industry professional), and sent a follow-up email with a suggested "80s" timeperiod bullet and got no respose. It didn't make it on the priority list. This is precisely the reason I agree that Stallman's concern is justified. Corporate marketing of the Linux history will continue to ignore the contributions of freedom-oriented pioneers unless customers and contributors complain.
Even if you feel (like I do) that Stallman's attempt to relabel things is impractical, feel free to encourage (nicely) corporate marketing to at least acknowledge the *significant* GNU contribution in their version of 'history' at suggest@redhat.com.
I'm almost positive that SCO has licensed rights for use of the term Unix from the Open Group or its precedessors. SCO bought UnixWare from Novell, Novell bought UnixWare from AT&T who originally owned the trademark for and SVR* implementation of UNIX. Novell owned the trademark and transferred it to the Open Group, if I recall correctly, while retaining the right to use it in their own product. I would imagine that that right was transferred to SCO with the sale of UnixWare.
UnixWare *would* violate the trademark if it weren't already licensed from the OpenGroup. But SCO almost certainly does have such a license.
Here are a few things I observed when visiting PEAR's labs and working a little with them as a engineering undergraduate at Princeton taking the human/machine interaction course that spent a week or so on PEAR-related stuff. These are impressions worn by time, rather than hard-core facts, but might give you a little better sense of where PEAR seems to be coming from:
1) They've been working on these experiments for well over a decade; billions of trials (each requiring a reasonably attentive human) are only made possible with significant amounts of time. This is not fly-by-night research. 2) From what I could tell, the research was made possible by two things: A) the university tenure system that allowed the engineering professor who switched fields to investigate this to stay employed (albeit shunted to the worst facilities down in the basement of the engineering building), and B) a small amount of funding from various U.S. defense/military research sources. 3) The fact that someone could keep this kind of research funded for two decades is itself remarkable. I found the reasoning used to justify NFS/DARPA grants memorable: if conciousness can effect sensitive electronic equipment, there is a big payoff in understanding the interactions of, say, fighter pilots with their $20 million dollar aircraft. Do such machines really break down more frequently under extreme operator stress, and why/how? Given the tens of billions spent on such equipment annually, spending tens, hundreds of thousands of that on research checking out these implications would be prudent. Not to mention the payoff if you do discover something truly new. That seems to be the primary argument for funding. 4) The (apocryphal?) story I heard about how this research got started went like this: Princeton undergrads have the opportunity to do independent research on topics of their choosing and most degrees require such a project to graduate. Jahn got interested in the research when one of his advisees, against his advice, did research into a form of this mind-affecting-reality phenomenon and he as advisor was unable to disprove the results or methodology. So he has been doing various experiments over time to replicate the results or try related methods. 5) The research has never to my knowledge satisfied the strong version of the skeptic's credo, "extraordinary claims demand extraordinary evidence", but the researchers have been claiming statistical significance for various findings for years. However, the statistical methods are subtle enough that I frankly find it hard to trust them. I've never recovered from my distrust of statistics after watching so many experts argue over the Monty Haul problem a few years back- there are just too many ways to overlook dependencies and definitional criteria. I suppose that if I found them reviewed in a peer-review statistics journal, I'd be able to muster a little more faith. On a related note, I'm not sure the skeptic's credo is really applicable to scientific judgements; perhaps over the long-haul its a useful measuring stick, but both valid and invalid scientific findings tend to start out with only modest amounts of evidence. 6) Regarding your specific descriptions of PEAR methodology. I would be concerned about the multiple-hypothesis apparatus you describe; I didn't see it in action but I intuitively wondered about its effects on the endeavor overall. However, if the hypothesis does give repeatable results, then it should be explainable and add to our store of knowledge. Repeatability is always the touchstone for this kind of research. Like you, this would probably be my biggest concern. As for other factors, there clearly was some labelling of people as talented and untalented within the project. I didn't hear of any "throwing out of data when the subjects feel tired, or are otherwise displeased with the results" but I wasn't involved in any real data collection so wouldn't know one way or the other. 7) I find it vaguely comforting that there is at least someone with strong previous academic experience (Jahn) is investigating these phenomenon. However, while I don't understand statistics well enough to trust their main assertions, it seems to me as a lay engineer that even if their work comes to naught, they were probably pushing the boundaries of knowledge about A) random number generators, developing a wide variety of physical sources of randomness necessary for their experiments, and B) practical application of statistics to such random/noise environments.
In summary, in an oddly perverse way it seems to me sort of like SETI-in-the-small: take a lot of noise and search for a signal. Weed out bad signals. Low potential for success, but high payoff if realized.
With DVDs and DMCA on the forefront, it's clear that cryptography has social ramifications that go beyond issues of security and privacy -- it's also a big technological stick companies hope to use to extract profits from consumers. Clearly this is true of the content companies (movies and audio recordings); more subtle are the motivations of the computer industry.
The computer industry players outwardly claim to want to make strong encryption available to everyone, since that is what their customers want to protect their privacy. And they don't want the hassle of multiple implementations with varying security levels and "foreign competition" (a buzzword that while theoretically true has not been substantiated with any real-life threat to revenues of the big companies complaining about it; small security-oriented companies I could sympathize with of course.) Overlooked however is the benefit that such computer companies, hardware and software, stand to gain: the ability to increase barriers to entry into their existing hardware/software markets through stronger protection of their intellectual property. The latest well-known example of this is Mattel/Learning Company's CyberPatrol encryption, but there are numerous other historic illustrations of this including Microsoft's infamous attempt to encrypt the Windows 3.1 beta error message routines indicating that DR-DOS was not supported.
Intel, for example, is already incorporating some encryption technology within their microprocessors (for currently reasonable purposes) that allows run-time modification of some of the chips' microcode for the purposes of issuing patches to bugs (e.g. the Pentium bug) on the chip. Details of this of course are only available under tightly restricted NDAs.
In the computer industry, you have to keep releasing new products to stay ahead of your competitors because they can, with time and some *legal* reverse-engineering, copy your product. Strong encryption technology built into hardware and software devices could make it practically near-impossible to perform such copying or legal reverse-engineering, thus raising even more substantial barriers to competition. Sure, protection of intellectual property is a good thing, handled reasonably well with existing legal mechanisms, (as seen by the industry's current success) but strong encryption would provide and cement a strong technological mechanism for protecting computer companies' (and others') intellectual property. The hardware or software code lying at the heart of a product can be encrypted whenever and wherever stored, and only unencrypted with the proper exchange of messages and keys.
This is a major social ramification of encryption that has remained unaddressed in the public at large, since the large computer companies (Intel, Microsoft, IBM, HP, etc.) best capable of implementing such complex mechanisms are not interested in advertising how much harder this will make the environment for their competition. Essentially, encryption greatly strengthens the technological capability of companies to arbitrarily "tie" two products together in a manner which competitors cannot substitute their product for either of the two components. While in the past, devices relied on a hard-to-copy physical interconnect to prevent copying or substitution of component parts, strong encryption enables a virtually foolproof virtual interconnect between components, and one that, unlike physical device specification, can make reverse-engineering nigh-impossible.
For example, take printer toner cartridges. If you can build build cheap, near-free encryption on both the toner cartridge and the printer, a printer designer can setup the printer to only work when a cartridge is inserted with the proper exchange of electronic encrypted messages. Because the encrypted message can involve a randomly-generated key component, an exchange-response protocol between the printer and the toner cartridge can become effectively impossible to reverse-engineer. Thus, someone who designs a printer cartridge that works in exactly the same manner physically will be prevented technologically from competing in the "printer consumables" market.
In that example, the printer cartridge is a solely physical component whose interoperability is limited solely by electronic encryption means. However, this encryption can similarly limit interoperability of computer hardware and software products in such a way that the encryption must be broken for an interoperable product to be designed.
Lest you think this concern is academic (computer companies haven't used weak encryption much, so why would they start using strong encryption now?), consider the singular difference between Microsoft's Kerberos implementation and standards-compliant ones. Microsoft fills one field with undocumented, and as I understand it, encrypted security ID information. (Better pointers to what is known about this undocumented feature appreciated.) In the name of selling you better security, the computer company sells you a self-serving product that locks you into their infrastructure and cycle of network externalities, limiting your ability to switch suppliers. (And in the case of Office2000 SR-1, it's not just a virtual network of products you have to connect with, but your system itself has to be *physically* connected to Microsoft over the Internet to register and activate their products.)
DMCA may have a loophole for interoperability purposes, but through strong encryption and techniques used in polymorphic viruses, computer companies should be able to insure that what isn't protected by law *is* protected by technology.
Remember, Bill Gates doesn't seem to think anti-trust law has any meaning or purpose in a world of software. On the contrary, it has more meaning than ever.
right, my point being that if your going to ooh-ah and speculate about Alpha vaporware, why examine it in light of the comparable Intel vaporware?
--LP
Re:Keep in mind- Intel is a moving target too
on
1.4-1.6 GHz Alphas
·
· Score: 2
I'm curious to see whether Intel designed the floating point units in Itanium or whether that was an HP contribution. If Intel got the HP floating point units as part of the joint-development deal, this would obviously narrow the gap substantially; HP and Alpha have been pretty neck and neck in the SPEC race the last few years.
--LP
Keep in mind- Intel is a moving target too
on
1.4-1.6 GHz Alphas
·
· Score: 4
This is good news for Alpha, but the delivery date will leave them under substantial pressure this fall and even after the new Alphas come out for two reasons:
Intel's Willamette will be out in Q3 running at 1.1 GHz to at least 1.4 GHz, coming pretty close or equal to these Alphas on megahertz, and likely on SPECint (integer and branching performance where Intel typically lags Alpha by 5-50%.) The proper comparison is current -products-to-current-products or future-at-dateX to future-at-dateX products, not current-to-future.
Alpha's pre-eminence in floating point is about to drop sharply due to Itanium and Willamette products from Intel with substantially improved floating point (many more fp execution units). Alpha has traditionally provided 3x the floating point of Intel. About 3-4 years ago, Intel realized that floating point was useful for the mass market (3D games) and the workstation space they were starting to enter, and these two processor cores are the first to reflect significant prioritization of floating point from Intel (SSE in Pentium III was a minor modification of a existing core design.)
While Intel has kept this mostly under wraps, some details have leaked out indicating that Willamette/Foster (Foster is the "Xeon" version of Willamette) will have equal or better floating point than Itanium, and Itanium's floating point has been publicly characterized in things like Kenneth's Intel roadmap. Note that the existence of only 8 registers on IA32 designs like Willamette probably substantially restrains performance on real apps. Intel will add 144 new SSE2 instructions according to Sharky and eliminate the switch time previously required when going from regular fp to SSE instructions. The memory bandwidth gap between Alpha and Intel will also narrow with Willamette/Itanium.)
Net: Intel may be about to catch up a significant amount on floating point, a historic Alpha differentiator, and Intel clock rates and integer/branch performance definitely keep pace at the 1-2 GHz levels.
--LP
Re:But is this really for the better?
on
Microsoft Loses
·
· Score: 2
Oh please. The stock has been bouncing between 90 and 115 for the last year or so. It was down in the upper 80s a week or two ago. The only people losing money are speculators who were betting on a settlement.
Either way, MHz is really a very poor metric for console performance.
Agreed. It was odd to see the reviewer revert to it when gauging his "excitement meter," despite having a generally decent grasp on other things.
A dedicated and optimized piece of hardware can often run at one third the speed of a normal CPU in MHz and still outperform it.
Note that the Emotion Engine DSP/VLIW/SIMD approach is *not* "dedicated and optimized" for a single purpose such as geometry calculations. It's got a lot of fp circuits lashed together in general-purpose form for AI, 3D geometry, or whatever other use software developers can think of. The GeForce geometry engine or the 3Dlabs "Gamma" geometry engine, in contrast, do contain dedicated ASIC circuits aimed at precisely those matrix operations needed in the geometry manipulation stages of the graphics pipeline. The Emotion Engine falls somewhere inbetween the pendulum of special purpose and general purpose circuitry, being both more specific than a CPU (i.e. due to lots of floating point circuitry, registers, and high-bandwidth memory paths), and more general and programmable than a standard geometry pipeline ASIC (i.e. with its DSP-like approach)
So does a more general CPU but a more specific geometry engine win (i.e. X-box) win out over a slower CPU with heavy floating point but still somewhat generalized geometry apparatus (PS2)? Of course! It's coming out 2 years later than PS2, what do you expect?
And I still think its all irrelevant in the end; how relevant can 16-66 Million triangles/sec (PS2 claims) or 300 M tris/sec (xbox claim) be when a worst-case NTSC 320x200 display at 24fps even with one triangle per pixel only requires 1.5 M triangles? Even factoring in overhead for theoretical vs. actual figures, the limited requirements of TV resolution makes console chip wars not too terribly relevant going forward, IMHO. (True, a best-case 640x480 TV res requires 7M tris/sec to hit one tri per pixel, which soaks up more cycles but even this still suggests that 300 M tris/sec is overkill/irrelevant.)
The technically-relevant bottlenecks for consoles are the display and network bandwidth, not the CPU and graphics!
A dozen function units hooked together can be really fast, but the logic for connecting them (both hardware and software) is the true determinant of how efficiently they can be used and thus the actual performance that can be attained. I find it hard to draw any conclusions about raw performance from the data described. Clearly the biggest thing Playstation benefits from is the miniscule screen resolution of TVs (640x480, 24fps max); how many games are there that aren't 24fps at 640x480 on the PC with practically any 3D card these days?
On a different note, I found it odd that the author shrugged his shoulders at a 600 MHz SIMD Intel processor (MMX/SSE = SIMD) in the X-box when speaking so glowingly of a 200 MHz SIMD one. I suppose the number of functional units differs, but it was still a little wierd given that the author focused on the MHz as being unimpressive. It seemed like he got caught in the understandable trap of looking at the X-box as a PC, not as a $300 console. For a console (and that *is* what we are talking about, right?) 600 Mhz would be a breakthrough, right?
Provocatively (?) yours, --LP
Re:Question for Iridium-knowledgeable, haiku form
on
A Eulogy for Iridium
·
· Score: 2
Ping time sucks, but while Quake players might care, for wireless phone webbrowsers (the actual target usage,) this isn't much of an issue.
It'd be odd to "guess rightly" that they have "plenty of bandwidth" when the backers claim that the lack of bandwidth was what made the system not worth keeping up. I'm curious exactly what the bandwidth characteristics of the satellites were.
And the ramifications if bandwidth was limiting or ping was the deal-killer are rather relevant. One implies that no satellite system at that height or higher would *ever* make sense for Internet access, the other implies simply that a properly designed system might be feasible.
Slashdot Mirror
(Note that Windows NT 4.0 didn't support Plug and Play either. A curious instance of slamming Linux for missing something that NT was also missing.)
Ironically, with Windows 2000, Microsoft's offerings have gotten worse in some dimensions; the hardware compatibility list has shrunk roughly in half.
The application story for Windows 2000 is similarly weaker since Microsoft has traded off backwards compatibility in an attempt to improve stability. A significant fraction of Windows NT 4.0 applications don't run under Windows 2000 due to new restrictions imposed to improve stability, for example, restricting applications' ability to write freely into the registry (oops, now there was a security/stability hole!).
--LP
It's a nice idea. It caused me to reflect on all the times I haven't posted on USENET or other places under my own name for precisely the reasons he's mentioned. Fear of my words coming back to haunt me (due to technical inaccuracy or being taken out of context, read by a future boss, etc.) has definitely led me to post less at times. Then again, maybe this is a good thing. ;-) But I agree; lack of a time horizon for computer-mediated communications definitely leads to self-censorship, something we should be wary of.
--LP
Did IRIX really start as a fresh implementation as the chart describes? I thought it was based off the BSD releases, based on early data sheets I have on IRIX 3.x...
--LP
If you want a mix of volunteer and commercial projects, try PopularPower</ a>. Their current project is optimizing flu vaccines, pretty cool imho.
--LP
If you're interested in JavaSpaces and distributed.net-type computing, you might check out http://www.popularpower.com/. It's hard to tell from the webpages whether it uses JavaSpaces per se, but it clearly uses Java virtual machines as the sandbox for distributing parallel computations across thousands of nodes on the Internet, and they do have Windows and Linux clients.
The current workload appears to be a pure volunteer effort aimed at helping develop better flu vaccines, but once installed, obviously other (for-pay) algorithms implemented in Java could be run on the client, with Java's sandbox preventing you from virus-type problems. Interesting stuff.
--LP
The other issue is that RAM prices have stabilized after huge drops from late 95 to early 99 that far exceeded the pace of Moore's law. RAM manufacturers were losing their shirts and not willing to double down investing in huge new factories back then that would be reducing the shortage now.
The pendulum shortages/crashes of commodity devices like RAM are one reason why Intel got out of the business in the late 1970s.
--LP
Disclaimer: I've tracked RAM prices on pricewatch for several years but my analysis is more conjecture based on reading trade press than first-hand knowledge.
I found it ironically funny that, checking DNS records, the owner of the obvious domain name for MS's latest techno-marketing initiative, dot.net, is a guy from Sun.
:)
Domain Name: dot.net
Administrative Contact, Technical Contact, Zone Contact: Comay, David S (DC115) David.Comay@ENG.SUN.COM
This must be another example of Microsoft "leadership" and "innovation" at work...
-LP (not connnected with Sun myself)
Hired most of the core guys who worked on DCE...
--LP
Yawn. What a disappointment. Here I wanted to see how quickly I could port my 3D "tetris" game to a cell phone and I have to wait another four months. This is a good reminder of how poor the quality of Slashdot information and Slashdot community information is. Well, here's one member of the community trying to get the facts I've learned out. We'll see if you moderators cooperate. :-) (Is my criticism of Slashdot too insightful for you? Oh, sorry, I'm not supposed to mention the invisible moderators in a plea for points, either overt or reverse- sychologically, right?)
Sarcastically yours,
--LP
Your first sentence is correct, your second is not. All products do benefit when they are compatibile and interoperable with one another. But Microsoft doesn't want to raise the value of all products. It only wants its own to benefit.
It makes perfect business sense for Microsoft to try to lower the value of competing products by preventing interoperability with its own. It's called lock-in, and it increases switching costs for users and barriers to entry for competitors. It's a strategy that makes perfect sense if you have a dominant (especially monopoly) position, and little or no sense if you don't have such a position. Harness network effects to exclusively benefit your product, what could be simpler?
The only time this doesn't pay off is if it sufficiently alienates customers or developers. So far, Microsoft has managed to hew a fine line where such alienation has not outweighed the benefits of its platform. It's up to knowledgeable people to point out the oft-hidden costs and risks of adopting Microsoft's technology approach.
Remember, in reality, most of Microsoft's succesful innovations have been *legal* innovations, beginning with their DOS contract and extending through various exclusive OEM agreements and their chiseling away at the Java contract and DOJ Consent Decree. Their trade-secret licensing of Kerberos and their attempt to license software on a renewal basis (first at universities) are just the latest examples of this. Just what you'd expect from a firm founded by the bright son of a lawyer.
--LinuxParanoid, paranoid for Linux's sake
...does not constitute contributory infringement if the product is widely used for legitimate, unobjectionable purposes...
This would seem to be the weakest argument of the Napster case; if the RIAA can show Napster is not widely used legitimately, do they win the case? There are legal Napster users out there (like myself, although I don't use Napster a lot), but I can't imagine there are too many of us.
--LP
Making available for public download is legally the same as publishing.
Hmm, while this rings true, I guess I'd have to look up a legal definition of publishing.
If I (a Napster user) never let anyone actually download, (a function supported by Napster that is not available in otherwise analogous FTP server software) then the premise of the statement (that the music on my hard disk is, in fact, available for public download) is false. Files on my hard disk do show up in Napster listings, but I haven't, as a matter of actual practice, had or let people download them.
It requires them to convince a judge that putting files on Napster constitutes publishing. I don't think they'll have any problems.
Perhaps. I suppose I would likewise view my point as picayune if either A) I'd never used Napster and relied on others to describe it to me, or B) I typically used it illegally. However, from the vantage point of a user attempting to take advantage of the software to move my (legal) music collection onto my hard disk, the distinction is highly relevant.
(I didn't find the DeCSS "legal excuse" that the software was for viewing DVDs on Linux rather than for cracking encryption and piracy very compelling myself, but then, that's partly because I wasn't part of the crowd trying to legally view DVDs on Linux.)
--LP
Not exactly. You're overlooking something. Metallica gave them a list of people who showed up in a list as potentially offering the song for download. *Offering*, not *pirating*. There's a difference, a crucial one for a (gasp) legal user of Napster such as myself. (I have used Napster to get recordings on my computer that I have already purchased in cassette form.)
Offering songs for download is not a crime. Actually transferring them is. It's quite possible (and has been true for me in my admittedly limited Napster usage) that one can "list" a copyrighted song without it ever being transferred, particularly if one likes obscure bands or does not stay online with Napster very long.
Napster also provides chat facilities and the ability to detect that someone is downloading your file and the ability to boot them. These enable me, the purchaser of the product, to enforce the copyright provisions that I have always been responsible for as with previous technologies by checking to see whether the recipient is pirating rather than getting a legal (fair use) copy from me. Given the way the Napster software scans my hard disk and does not offer me an up-front choice of exactly which songs to list, this is a quite-likely possibility for someone with a mix of free (i.e. Grateful Dead) recordings mixed with a list of non-free ones who legitimately and legally wants to use Napster to share music while storing his/her own music on their computer.
To accurately and properly accuse 335,435 users of pirating (breaking copyright), Metallica would actually have to demonstrate that each one of them had not just offered a file, but transferred it to someone else illegally (in the process, verifying that said file was, in fact, a Metallica song for which they held the copyright.)
So I see this as one big PR move in an attempt to intimidate Napster and Napster users. Those 335,435 names (handles, actually) are nothing more than a huge list of "subversives" using the "wrong" software, not a list of 335,435 criminals. Innocent until proven guilty (in the US), right?
Now, the *real* problem with the Napster approach: But just because a legal loophole may knock this issue off the table, don't get arrogant. The labels can subvert Napster whenever they want by flooding Napster with falsely-named files, if they ever get enough clue and guts to do so (hiding their fingerprints to avoid backlash by pointing to those anarchist internet trolls as the culprits.) For example, instead of a real track, put a 30-second promo clip appended with 3-5 minutes of white noise. There are countermeasures to this, but ultimately the only real solution to limiting unjust corporate power is for citizens to either stop buying or organize politically and press for *real* "fair use" copyright provisions.
--LP
...making those MP3s available for public download (what you call trading) happens to be a copyright violation...
Close, but technically false. Making them available for download is not a crime. Actually transferring them is. It's quite possible (and has been true for me in my Napster usage) that one can "list" a copyrighted song without it ever being transferred, particularly if one likes obscure bands or does not stay online with Napster very long.
For Metallica to claim I broke their copyright requires them to show I transferred a file to a person who did not own a legit copy. To have a real legal case (AFAICT, IANAL), they would have to mount a sting for each person so accused: attempt to actually download a file from me (and confirm that the file really was "their" content).
--LP
Note how Red Hat ignores the contributions of GNU (the GPL, gcc, core UNIX tools) in their version "Linux History." For someone who watched and played with Linux in 1992-1993, this ignores the major way in which GNU tools enabled Linus and collaborators worldwide to develop Linux. I've talked to the marketing manager personally about this (as an industry professional), and sent a follow-up email with a suggested "80s" timeperiod bullet and got no respose. It didn't make it on the priority list. This is precisely the reason I agree that Stallman's concern is justified. Corporate marketing of the Linux history will continue to ignore the contributions of freedom-oriented pioneers unless customers and contributors complain.
Even if you feel (like I do) that Stallman's attempt to relabel things is impractical, feel free to encourage (nicely) corporate marketing to at least acknowledge the *significant* GNU contribution in their version of 'history' at suggest@redhat.com.
--LP
I'm almost positive that SCO has licensed rights for use of the term Unix from the Open Group or its precedessors. SCO bought UnixWare from Novell, Novell bought UnixWare from AT&T who originally owned the trademark for and SVR* implementation of UNIX. Novell owned the trademark and transferred it to the Open Group, if I recall correctly, while retaining the right to use it in their own product. I would imagine that that right was transferred to SCO with the sale of UnixWare.
UnixWare *would* violate the trademark if it weren't already licensed from the OpenGroup. But SCO almost certainly does have such a license.
--LP
Here are a few things I observed when visiting PEAR's labs and working a little with them as a engineering undergraduate at Princeton taking the human/machine interaction course that spent a week or so on PEAR-related stuff. These are impressions worn by time, rather than hard-core facts, but might give you a little better sense of where PEAR seems to be coming from:
1) They've been working on these experiments for well over a decade; billions of trials (each requiring a reasonably attentive human) are only made possible with significant amounts of time. This is not fly-by-night research.
2) From what I could tell, the research was made possible by two things: A) the university tenure system that allowed the engineering professor who switched fields to investigate this to stay employed (albeit shunted to the worst facilities down in the basement of the engineering building), and B) a small amount of funding from various U.S. defense/military research sources.
3) The fact that someone could keep this kind of research funded for two decades is itself remarkable. I found the reasoning used to justify NFS/DARPA grants memorable: if conciousness can effect sensitive electronic equipment, there is a big payoff in understanding the interactions of, say, fighter pilots with their $20 million dollar aircraft. Do such machines really break down more frequently under extreme operator stress, and why/how? Given the tens of billions spent on such equipment annually, spending tens, hundreds of thousands of that on research checking out these implications would be prudent. Not to mention the payoff if you do discover something truly new. That seems to be the primary argument for funding.
4) The (apocryphal?) story I heard about how this research got started went like this: Princeton undergrads have the opportunity to do independent research on topics of their choosing and most degrees require such a project to graduate. Jahn got interested in the research when one of his advisees, against his advice, did research into a form of this mind-affecting-reality phenomenon and he as advisor was unable to disprove the results or methodology. So he has been doing various experiments over time to replicate the results or try related methods.
5) The research has never to my knowledge satisfied the strong version of the skeptic's credo, "extraordinary claims demand extraordinary evidence", but the researchers have been claiming statistical significance for various findings for years. However, the statistical methods are subtle enough that I frankly find it hard to trust them. I've never recovered from my distrust of statistics after watching so many experts argue over the Monty Haul problem a few years back- there are just too many ways to overlook dependencies and definitional criteria. I suppose that if I found them reviewed in a peer-review statistics journal, I'd be able to muster a little more faith. On a related note, I'm not sure the skeptic's credo is really applicable to scientific judgements; perhaps over the long-haul its a useful measuring stick, but both valid and invalid scientific findings tend to start out with only modest amounts of evidence.
6) Regarding your specific descriptions of PEAR methodology. I would be concerned about the multiple-hypothesis apparatus you describe; I didn't see it in action but I intuitively wondered about its effects on the endeavor overall. However, if the hypothesis does give repeatable results, then it should be explainable and add to our store of knowledge. Repeatability is always the touchstone for this kind of research. Like you, this would probably be my biggest concern. As for other factors, there clearly was some labelling of people as talented and untalented within the project. I didn't hear of any "throwing out of data when the subjects feel tired, or are otherwise displeased with the results" but I wasn't involved in any real data collection so wouldn't know one way or the other.
7) I find it vaguely comforting that there is at least someone with strong previous academic experience (Jahn) is investigating these phenomenon. However, while I don't understand statistics well enough to trust their main assertions, it seems to me as a lay engineer that even if their work comes to naught, they were probably pushing the boundaries of knowledge about A) random number generators, developing a wide variety of physical sources of randomness necessary for their experiments, and B) practical application of statistics to such random/noise environments.
In summary, in an oddly perverse way it seems to me sort of like SETI-in-the-small: take a lot of noise and search for a signal. Weed out bad signals. Low potential for success, but high payoff if realized.
--LP
With DVDs and DMCA on the forefront, it's clear that cryptography has social ramifications that go beyond issues of security and privacy -- it's also a big technological stick companies hope to use to extract profits from consumers. Clearly this is true of the content companies (movies and audio recordings); more subtle are the motivations of the computer industry.
The computer industry players outwardly claim to want to make strong encryption available to everyone, since that is what their customers want to protect their privacy. And they don't want the hassle of multiple implementations with varying security levels and "foreign competition" (a buzzword that while theoretically true has not been substantiated with any real-life threat to revenues of the big companies complaining about it; small security-oriented companies I could sympathize with of course.) Overlooked however is the benefit that such computer companies, hardware and software, stand to gain: the ability to increase barriers to entry into their existing hardware/software markets through stronger protection of their intellectual property. The latest well-known example of this is Mattel/Learning Company's CyberPatrol encryption, but there are numerous other historic illustrations of this including Microsoft's infamous attempt to encrypt the Windows 3.1 beta error message routines indicating that DR-DOS was not supported.
Intel, for example, is already incorporating some encryption technology within their microprocessors (for currently reasonable purposes) that allows run-time modification of some of the chips' microcode for the purposes of issuing patches to bugs (e.g. the Pentium bug) on the chip. Details of this of course are only available under tightly restricted NDAs.
In the computer industry, you have to keep releasing new products to stay ahead of your competitors because they can, with time and some *legal* reverse-engineering, copy your product. Strong encryption technology built into hardware and software devices could make it practically near-impossible to perform such copying or legal reverse-engineering, thus raising even more substantial barriers to competition. Sure, protection of intellectual property is a good thing, handled reasonably well with existing legal mechanisms, (as seen by the industry's current success) but strong encryption would provide and cement a strong technological mechanism for protecting computer companies' (and others') intellectual property. The hardware or software code lying at the heart of a product can be encrypted whenever and wherever stored, and only unencrypted with the proper exchange of messages and keys.
This is a major social ramification of encryption that has remained unaddressed in the public at large, since the large computer companies (Intel, Microsoft, IBM, HP, etc.) best capable of implementing such complex mechanisms are not interested in advertising how much harder this will make the environment for their competition. Essentially, encryption greatly strengthens the technological capability of companies to arbitrarily "tie" two products together in a manner which competitors cannot substitute their product for either of the two components. While in the past, devices relied on a hard-to-copy physical interconnect to prevent copying or substitution of component parts, strong encryption enables a virtually foolproof virtual interconnect between components, and one that, unlike physical device specification, can make reverse-engineering nigh-impossible.
For example, take printer toner cartridges. If you can build build cheap, near-free encryption on both the toner cartridge and the printer, a printer designer can setup the printer to only work when a cartridge is inserted with the proper exchange of electronic encrypted messages. Because the encrypted message can involve a randomly-generated key component, an exchange-response protocol between the printer and the toner cartridge can become effectively impossible to reverse-engineer. Thus, someone who designs a printer cartridge that works in exactly the same manner physically will be prevented technologically from competing in the "printer consumables" market.
In that example, the printer cartridge is a solely physical component whose interoperability is limited solely by electronic encryption means. However, this encryption can similarly limit interoperability of computer hardware and software products in such a way that the encryption must be broken for an interoperable product to be designed.
Lest you think this concern is academic (computer companies haven't used weak encryption much, so why would they start using strong encryption now?), consider the singular difference between Microsoft's Kerberos implementation and standards-compliant ones. Microsoft fills one field with undocumented, and as I understand it, encrypted security ID information. (Better pointers to what is known about this undocumented feature appreciated.) In the name of selling you better security, the computer company sells you a self-serving product that locks you into their infrastructure and cycle of network externalities, limiting your ability to switch suppliers. (And in the case of Office2000 SR-1, it's not just a virtual network of products you have to connect with, but your system itself has to be *physically* connected to Microsoft over the Internet to register and activate their products.)
DMCA may have a loophole for interoperability purposes, but through strong encryption and techniques used in polymorphic viruses, computer companies should be able to insure that what isn't protected by law *is* protected by technology.
Remember, Bill Gates doesn't seem to think anti-trust law has any meaning or purpose in a world of software. On the contrary, it has more meaning than ever.
--LinuxParanoid, paranoid for Linux's sake
right, my point being that if your going to ooh-ah and speculate about Alpha vaporware, why examine it in light of the comparable Intel vaporware?
--LP
I'm curious to see whether Intel designed the floating point units in Itanium or whether that was an HP contribution. If Intel got the HP floating point units as part of the joint-development deal, this would obviously narrow the gap substantially; HP and Alpha have been pretty neck and neck in the SPEC race the last few years.
--LP
Net: Intel may be about to catch up a significant amount on floating point, a historic Alpha differentiator, and Intel clock rates and integer/branch performance definitely keep pace at the 1-2 GHz levels.
--LP
Oh please. The stock has been bouncing between 90 and 115 for the last year or so. It was down in the upper 80s a week or two ago. The only people losing money are speculators who were betting on a settlement.
Wah.
--LP
Agreed. It was odd to see the reviewer revert to it when gauging his "excitement meter," despite having a generally decent grasp on other things.
A dedicated and optimized piece of hardware can often run at one third the speed of a normal CPU in MHz and still outperform it.
Note that the Emotion Engine DSP/VLIW/SIMD approach is *not* "dedicated and optimized" for a single purpose such as geometry calculations. It's got a lot of fp circuits lashed together in general-purpose form for AI, 3D geometry, or whatever other use software developers can think of. The GeForce geometry engine or the 3Dlabs "Gamma" geometry engine, in contrast, do contain dedicated ASIC circuits aimed at precisely those matrix operations needed in the geometry manipulation stages of the graphics pipeline. The Emotion Engine falls somewhere inbetween the pendulum of special purpose and general purpose circuitry, being both more specific than a CPU (i.e. due to lots of floating point circuitry, registers, and high-bandwidth memory paths), and more general and programmable than a standard geometry pipeline ASIC (i.e. with its DSP-like approach)
So does a more general CPU but a more specific geometry engine win (i.e. X-box) win out over a slower CPU with heavy floating point but still somewhat generalized geometry apparatus (PS2)? Of course! It's coming out 2 years later than PS2, what do you expect?
And I still think its all irrelevant in the end; how relevant can 16-66 Million triangles/sec (PS2 claims) or 300 M tris/sec (xbox claim) be when a worst-case NTSC 320x200 display at 24fps even with one triangle per pixel only requires 1.5 M triangles? Even factoring in overhead for theoretical vs. actual figures, the limited requirements of TV resolution makes console chip wars not too terribly relevant going forward, IMHO. (True, a best-case 640x480 TV res requires 7M tris/sec to hit one tri per pixel, which soaks up more cycles but even this still suggests that 300 M tris/sec is overkill/irrelevant.)
The technically-relevant bottlenecks for consoles are the display and network bandwidth, not the CPU and graphics!
--LP
A dozen function units hooked together can be really fast, but the logic for connecting them (both hardware and software) is the true determinant of how efficiently they can be used and thus the actual performance that can be attained. I find it hard to draw any conclusions about raw performance from the data described. Clearly the biggest thing Playstation benefits from is the miniscule screen resolution of TVs (640x480, 24fps max); how many games are there that aren't 24fps at 640x480 on the PC with practically any 3D card these days?
On a different note, I found it odd that the author shrugged his shoulders at a 600 MHz SIMD Intel processor (MMX/SSE = SIMD) in the X-box when speaking so glowingly of a 200 MHz SIMD one. I suppose the number of functional units differs, but it was still a little wierd given that the author focused on the MHz as being unimpressive. It seemed like he got caught in the understandable trap of looking at the X-box as a PC, not as a $300 console. For a console (and that *is* what we are talking about, right?) 600 Mhz would be a breakthrough, right?
Provocatively (?) yours,
--LP
Ping time sucks, but while Quake players might care, for wireless phone webbrowsers (the actual target usage,) this isn't much of an issue.
It'd be odd to "guess rightly" that they have "plenty of bandwidth" when the backers claim that the lack of bandwidth was what made the system not worth keeping up. I'm curious exactly what the bandwidth characteristics of the satellites were.
And the ramifications if bandwidth was limiting or ping was the deal-killer are rather relevant. One implies that no satellite system at that height or higher would *ever* make sense for Internet access, the other implies simply that a properly designed system might be feasible.
--LP