Slashdot Mirror
Slashback: Hilbert's, Transgenic, Silicon
Still an acorn at this point. Jose Nazario writes with a correction to my recent post claiming that OpenBSD had gained a "fuzzy" user-profiling IDS. Jose writes: "It is NOT in tree. it is a privately developed research project. It is not an official project."
And Yes, the Apple I schematics were available, too. In response to the recent article about the freely available chip design from opencores.org implemented by Flextronics, Henry Keultjes offers a reminder that this is not the first time chip whose internals have been open for inspection:
"Happened quite some time ago with PowerPC. That's the essence of Microsoft's deal with IBM because without that Open Architecture Microsoft would have had to buy a lot more than it did. This for example is used in a roughly $150 French set-top box that has USB and, according to a friend in the UK who has tried that, runs just fine as a PC with the attached USB HDD, KB and rodent."
Could Wayne Inouye sell you an eMachine? After reading many pointed comments in the story about eMachine's Athlon offerings, arrasmith writes "To add to the topic of AMD64 eMachines and the launch of "I hate eMachine" posts I'll throw out why you should buy one.
eMachines are the number-3 seller of computers, only behind Dell and HP. If you are wondering about how that happened, you need to read about the new CEO.
Wayne Inouye has had some articles published about him in Business Week and Forbes. Great articles on how you can sell good computers at reasonable prices. And if you are wondering why eMachines is selling an AMD64 system read the Business Week article."
OK, as long as you buy it from us. Alien54 writes "As reported in the most recent Spyware Info Newsletter, Dell seems to have listened to the criticism handed to them last week, after their decision to forbid tech support persons from providing assistance to spyware-infected customers became public knowledge. They have partnered with PestPatrol, Inc. to sell Pest Patrol's spyware removal software to Dell customers. It is interesting to note that Dell does not recommend any freeware or shareware product because 'we cannot test these open source utilities reliably.' Which is simply silly, of course."
Utah may not be Utopia after all. brysnot writes "The Salt Lake Tribune reports that the Utopia project, which plans to run fiber to every home in Utah, has miscalculated its 2003 budget and now needs each member cities to come up with an additional $250,000. Also reported is that 'Its largest member, Salt Lake City, is uncertain whether to provide financial backing to guarantee payment of the principal and interest on the bonds the project needs -- a development that could force the project to be scaled back.'"
Writes Lighthop "The best way to overcome Qwest's vast resources and well orchestrated opposition is for citizens and business owners to speak out and let their city council members know we support them in approving UTOPIA's funding. We have to be visible and give them some political cover.
The 18 UTOPIA member cities are Brigham City, Cedar City, Cedar Hills, Centerville, Layton, Lindon, Midvale, Murray, Orem, Payson, Perry, Riverton, Roy, Salt Lake City, South Jordan, Taylorsville, Tremonton and West Valley."
Hilbert's 16th is still a problem. commodoresloat writes "The work of Elin Oxenhielm, the 22-year old Swedish student who apparently solved part of the 16th Hilbert problem, is coming under heavy fire from some prominent mathematicians, including her own adviser, who said the work contained "serious mistakes, which I think any educated mathematician can easily see." Here's an article in English. Oxenhielm responded to the criticism by saying that the journal that accepted her work, which now owns the copyright, is responsible for any errors. More information on this weblog."
Periscope is up, showdown commences. McSpew writes "The Register states that Microsoft's patents on the FAT filesystem may be subject to new scrutiny, thanks to their announced plan to collect royalties from media and CE manufacturers. The Public Patent Foundation is behind the effort to get the USPTO to start from scratch with Microsoft's FAT patents."
FDA gives GM fish sales the eerie green light. fishfishfish writes "The US Food and Drug Administration (FDA) Tuesday released a statement saying that it will not be stopping the sale of transgenic Zebra danios in the USA. The move could allow fish retailers in any U.S. state to sell the fish. Apart from California, where Arnie has banned them..."
From the weblog:
The fact is, though, that Zhou was an advisor for Oxenhielm's masters degree. She is neither her professor nor her current advisor.
o, assuming I get hold of one of these AMD 64-bit boxes, how hard/easy is it to get Linux compiled for 64-bit. What are the pitfalls with gcc (is an int 64 bit in 64-bit mode ?)
Here is a list of supported distros. And yes, I believe an int is 64 bits in 64 bit mode
"We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
As one who has refereed math papers, I think that this is not true. When I am sent a letter asking me to referee, I am asked to comment on how important the result is, and I am asked to assess how correct the paper is, but often I am explicitly told that errors in the paper are the responsibility of the author, and not the referee.
According to the SLTrib article:
"...needs member cities to pony up an additional $250,000 so it can continue to pursue its bond offering."
So it looks like they're just $250k short, not $4.5 million short as the poster seemed to indicate. In fact, if I'm reading this right, it means each city would only need to come up with ~$14k each, if they're going to split it equally.
... And so it comes to this.
the journal owns the text, not what the text says.
big difference.
In "ordinary errors" should have been catched in the review process. If not, the journal has bad reviewers.
It sometimes cost money to publish scientific articles.
In essance, you pay them to review your paper.
Major errors that slip through, should be blamed on the journal, yes.
Warning: This sig contains a small bug. ==> *
(1) Uh.. can you even "sell" the copyright of mathematical proofs like this? Mathematics is truly in a scary and sorry state if you have to hand over full copyright of such a work to get it published...
You can't copyright a mathematical proof, but you can copyright an article in which a proof is given. In accademia there is nothing unusual about having to hand over the copyright to your work in order to get it published. In some fields accademics can't even give their work away - they have to pay journals to publish their articles. No big surprise either. If you ever spend any time reading the crap that gets published you will soon wish that you were being paid to read it.
As was pointed out numerous times in the original slashdot article, the patents refer to long file names. If you don't implement them, then no problem, so why insist on saying the patent is on FAT?
I have always found that directly e-mailing the author of any paper buried like this always results in them getting a copy to me. Academics are great about stuff like that. Wonderful people. I'll never forget asking for a copy of one wavelet paper from a researcher at an Italian University and three days later this enormous box full of copies of every paper the author had written turned up on my doorstep - and I don't even live on the same continent as Italy.
No, on 64-bit Linux platforms int is still 32 bits, while long is 64 bits. Pointers are also 64 bits.
The two most common C models are commonly referred to as ILP32 (int, long, pointers all 32 bit) or LP64 (long and pointer are 64 bit).
It's called "Submarining a patent." Read Title III.
The real problem is how difficult it is to define whether a product has become ubiquitous. For a lesson in how difficult that is, refer to CompuServe's superficially compelling arguments about the dominance of JPEG that allowed them to fool a judge into thinking the resurfacing of the LZ patents was okay. Sometimes a patent really can't be judged in time, and sometimes a company gets into commercialization beforehand knowing fully well that it'll have to stop; see the issue with the chemical that made wacky wall walkers, and Klutz Press.
StoneCypher is Full of BS
What do you need? What will it cost you?
I set up an uncompressed Knoppix on a dual-boot for an eMachine dial-up user new to linux. Didn't go badly. 64 meg video was OK. Response was OK. There was a proprietary modem driver available with a crippled demo download that installed fine. If you just need a computer and can get a good price, I wouldn't knock it.
If your code cares beyond that, use the standard C types (defined in stdint.h) that specify the sizes.
From the many year old C99 standard (part B.17 Integer types ):
int was defined in K&R1 to the best size for the CPU to deal with. Short and char could be shorter, long could be longer. Then they had a chart showing how several "common" machines implimented it, which included one machine that implimented all of the above types with 36 bits.
I still think that int should be the easiest size for the machine to deal with. If your intiger math is all 64 bits, when I say int I mean that I don't want you take an extra step to make the result fit into 32 bits. If your CPU is 16 bits, then by int I mean don't go through the extra effort to do 32 bit addition.
You can find a technical description Here, however.
In general, most Unixes and Linux (as you say) have adopted the LP64 model where longs/pointers are 64-bits and ints are 32 bits (some gory details here. (Cray's Unix is an exception; it's ILP64).
Windows OSes however have adopted the LLP64 model where ints and longs are 32-bits still, but long longs and pointers are 64-bits (gory Windows details here and here.)
Both 32-bit Windows and Unix traditionally used ILP32, so the porting characteristics moving to 64-bit code are slightly different across the two platforms.
--LinuxParanoid
When you are talking about different processor architectures, distro does matter as it provides the precompiled binaries for your system to run. A program compiled for a PowerPC won't run on a IA32 (aka i386) system and vice versa.
The kernel and modules are even more critical. Sometimes the basic design of computers using the same processor are different, so the kernel/bootloader has to use different code to load the system. The only examples I can think of right now might be something like the Amiga, the classic Mac, and Atari ST. They all used the 680x0 processor, but I believe the rest of their designs were completely different.
It also matters which hardware options were precompiled into your kernel. If the distro left out ISA support (perhaps they assume everyone uses only PCI), then you can't boot their system on an older computer which uses ISA cards--such as a 386 or many 486s.
Even source based distros (such as Gentoo) have to use binaries at some point. You can't compile without a compiler binary, and you can't run a compiler binary (or even boot) without a kernel binary. Though with source based the maintainers probably have less work to support each new processor.
So, yes it does make a big difference which distro you use when you want to use a new (or less common or just different from what they decided to support) processor architecture. Most of them support the most common type (IA32), but some distros may wait before they try to support these new 64 bit processors, if at all. However IA32 support should work fine--assuming they are fully backwards compatible. The software just won't take advantage of the extra power.
In mathematics, the accuracy of a proof is the responsibility of the author. A referee will attempt to determine the correctness of a proof but neither an editor nor a referee is ultimately responsible. Publishing an incorrect proof is not always bad; the "Yamabe conjecture" arose from a paper by Yamabe in 1960 (Osaka Math. Journal, Vol. 12, pp. 21-37) which was accepted as correct. (Rick Schoen provided a correct proof for the case of compact manifolds in 1984 and, for example, Zhiren Jin provided a counterexample for noncompact manifolds in 1986.) However, claiming that the publisher is responsible for errors is silly and unprofessional.
You can get Goedel's most famous work in paperback from Dover. His complete works are available in hardcover. Cohen's work was published four times, the latest in 1998. You can get a decent collection of papers by Tarski in print for $30.
If you want to see Cohen's proof, there are more modern (post-modern?) books on set theory with simplified treatments.
(disclaimer: my background in dynamical systems, much less this particular problem, is not that strong)
...."
:)
The second part of Hilbert's 16th problem deals with limit cycles, the way things will go on eventually in dynamical systems if they are not disturbed externally. The subproblem 2/3 of this problem (it's the indexing that makes math complicated..) asks if there exists an upper bound on the number of different limit cycles one can have in the system.
Oxenhielm attacks the problem by considering first a special case called the Lienard equation and approximating its solution by harmonic oscillation. The proof begins: "Noticing that the state variable x of the Lienard equation (1) behaves approximately like a sine function in simulations (see Fig.1),we assume -- in order to make a good approximation of x -- that both state variables are dominated by a harmonic term
Now, to my engineer's eyes, the functions in Fig.1 seem more like triangular waves, with definitely more than one single frequency component. Yet the accuracy of the approximation has not been considered at all in the paper. Also, 'proof by looking at results of simulations' is not really valid if you don't have any other evidence.
Another bad part is on page 6, where it is claimed that "Note that the method of describing functions may be used in a similar manner as in the proof above,to find the upper bounds for the Hilbert number in any planar polynomial vectorfield. Thus, it is possible to completely solve the second part of Hilbert's 16th problem by using this approach."
Wait a minute, how did that happen? What if the harmonic approximation fails on other than Lienard equations? It might just work, I have no idea, but this assertment hardly proves the fact.
Note however that this is very different from Andrew Wiles' proof of the Fermat conjencture. While very few people in the world could understand the odd-hundred pages of Wiles' proof, Oxenhielm's paper is just eight pages of much more accessible mathematics.
But I have a paper in the review process myself, and sure as hell would hate to see nonqualified people discussing its validity publicly, so maybe I'll just shut up now
You were joking but you are not far off. We scientists do much research with public and charity money. We then submit a paper, must pay some of the costs of the printing, and then to read it must subscribe to the journal. (OK, some journals are free). There are several campains going on in the scientific community to get all journals for free, but this is still far off. , ..my 0.02EUR anyway
There are many great papers that I would like to read and must resort to
a) writing to the first author and asking for a copy,
b) get a copy for the British Library for around 20EUR,
c)buy a single copy from the publisher at around 30EUR each paper.
I host on my website some of my papers and am breaking copyright law by doing so. THe way I see it I have no choice. If some young student wants to work in my lab and they want to see what we do and to study before the interview how else can they read this work.