Slashdot Mirror

...here, quickly improved it to 0.5 and 1 km, and then 10 km. Don't quite know why Nature thought this particular paper was so revolutionary -- wake me when they get to about 300 km, the minimal bounce-off-satellite trip.

Bob Lucky is one of the wisest engineers on the planet. Read on:

http://www.lanl.gov/projects/sfc/96/bios/lucky.htm l

Robert W. Lucky is Corporate Vice President of Applied Research at Bellcore.

He was born in Pittsburgh, Pa., and attended Purdue University, where he received a B.S. degree in electrical engineering in 1957, and M.S. and Ph.D. degrees in 1959 and 1961. After graduation he joined AT&T Bell Laboratories in Holmdel, NJ, where he was initially involved in studying ways of sending digital information over telephone lines. The best known outcome of this work was his invention of the adaptive equalizer - a technique for correcting distortion in telephone signals which is used in all high speed data transmission today. The textbook on data communications which he co-authored became the most cited reference in the communications field over the period of a decade.

At Bell Labs he moved through a number of levels to become Executive Director of the Communications Sciences Research Division in 1982, where he was responsible for research on the methods and technologies for future communication systems. In 1992 he left Bell Labs to assume his present position at Bellcore.

He has been active in professional activities, and has served as President of the Communications Society of the IEEE (Institute of Electrical and Electronics Engineers), and as Vice President and Executive Vice President of the parent IEEE itself. He has been editor of several technical journals, including the Proceedings of the IEEE, and since 1982 he has written the bi-monthly "Reflections" column of personalized observations about the engineering profession in Spectrum magazine. In 1993 these "Reflections" columns were collected in the IEEE Press book Lucky Strikes...Again.

Dr. Lucky is a Fellow of the IEEE and a member of the National Academy of Engineering. He is also a consulting editor for a series of books on communications through Plenum Press. He has been on the advisory boards or committees of many universities and government organizations, and was Chairman of the Scientific Advisory Board of the United States Air Force from 1986-1989. He was the 1987 recipient of the prestigious Marconi Prize for his contributions to data communications, and has been awarded honorary doctorates from Purdue University and the New Jersey Institute of Technology. He has also been awarded the Edison Medal of the IEEE and the Exceptional Civilian Contributions Medal of the U.S. Air Force.

Dr. Lucky is a frequent speaker before both scientific and general audiences. He has been an invited lecturer at about one hundred different universities, and has been the guest on a number of network television shows, including Bill Moyers' "A World of Ideas," where he has discussed the impacts of future technological advances. He is the author of the popular book Silicon Dreams, which is a semi-technical and philosophical discussion of the ways in which both humans and computers deal with information.

Dr. Lucky and his wife, Joan, currently reside in Fair Haven, NJ.

It is untested as of now. It is based purely on theoretical reasoning. If anyone wants to try and it test it in comparison to other techniques, I'd be very interested in hearing the outcome.

I'm not sure why this particular article needed to be posted, as it's just one of several alternative approaches and an untested one at that. On Paul's page, he also lists several published academic papers with other alternatives -- all actually tested, of course.

Gary is basically right in questioning the use of the word "Bayesian". Paul's approach is more about weighing "evidence" as given by the appearance of certain words, rather than in figuring out the probability of spam assuming a "prior". See Paul's explanation, but if you check the article he references at the end, you'll note that the method Paul uses is only one of several methods to solve an underspecified problems. It's a reasonable guess, not necessarily the only guess.

Looking at another article Paul references, given the word independence assumption, the more formal Naive Bayesian approach calculates as follows:
p(spam) = [ p(spam)*p(word1|spam)*...*p(wordn|spam) ] / [ p(spam)*p(word1|spam)*...*p(wordn|spam) + p(!spam)*p(word1|!spam)*...*p(wordn|!spam)]

This is similar to Paul's approach except for including a "prior" assumption of p(spam) -- the expected probability of any email being spam, calcuated from the historically observed frequency of spam. By leaving it out, Paul implicitly assumes that 50% of mail is spam -- that's his "prior" estimate of the spam rate. Given the other adjustments he makes to his sample, that appears to be acceptable in practice. (Paul overweights the spam prior, but also overweights the effects of "good" words.)

I'd personally prefer to overweight the "good" e-mails entirely rather than just put a "good-multiplier" on them like Paul does, but that's just quibbling over small bits.

As to the bit that Gary raises about Paul assuming a spam probability for an unknown word -- Paul originally said .2, then revised to .4, but really should have put it at .5 or just excluded it from all calculations. A new word has no robustness as a predictor (which is why Paul dropped words that didn't appear five times anyway). In practice, a new word at .4 isn't going to be among the 15 most interesting words to make the calculation from, anyway.

-XDG

Why not just use LinuxBIOS? A minimal footprint system should fit onto a reasonable (32MB) flash part.

Most BIOSes are designed to fit onto 128k or 256k parts. No real reason for this limit other than fact that these parts in volume are pretty cheap and they do the job required.

A company I used to work for made some embedded devices where we put the BIOS, OS and applications all on a 32MB flash part. It was basically an i386 platform with some custom hardware and software. Heck it could even run Win9x off a disk drive.

Seeing as how the Linux kernel replaces most of the functionality of the BIOS will setting any of these options really make a difference?

Any kernel developers out there care to chime in?

just to add more confusion to the story... one can show that the casimir force between two objects can also be viewed as a van der waals force. imagine two dielectrics separated by vacuum. in the casimir picture, the two dielectrics will attract because of vacuum zero-point energy fluctuations. as someone else pointed out earlier, this is, in very simple terms, because certain frequencies are not allowed in the vacuum creating a pressure. on the other hand, one can also generate the same force by having the particles in the dielectricum fluctuate, i.e. generate van der waals forces due to "dipoles".

from a physics point of view one has a hamiltonian (function which describes the whole system) for which one can integrate out the degrees of freedom of the dielectric or the degrees of freedom of the electromagnetic vacuum. if you remove the dielectric degrees of freedom, you get equations for the EM field which will give you a casimir force when you look at the cero point energies of the system with and without the dielectricum. if you integrate out the field, you will get an effective interaction term for the dielectrica and hence a vdW force.

note also, that both, van der waals and casimir forces depend with a power law on the distance between the objects, usually at a high power and are therefore extremely short-ranged. there are some nice papers and experiments measuring both at the usual archive xxx.lanl.gov. just search for casimir...

The national labs typically have the latest equipment, best training, and real job security.
Sandia and Los Alamos are great examples of this.
However, for these 2 you will have to LIVE IN NEW MEXICO.

(your living standards may vary, I just like "real" cities)

Sandia

Los Alamos

Here is an expository article from the Journal of the AMS about the Langlands program. Results of Lafforgue are used to prove some very nice theorems.

Here is a link to an article by Lafforgue in Inventiones Mathematicae, one of the world's most prestigious mathematics Journals. Malheursement, cet article est en français.

Here is the Mathematical Reviews citation for the Lafforgue paper. You can browse the articles cited by him.

Also, if anyone is interested, here is a paper by Voevodsky about some of his work in motivic cohomology.

I truly take pride in this discovery... mostly because I attend UW. But I suppose a love of physics helps in that area, too.

Anyways, here's a somewhat technical article regarding the research (PDF).

Oh, and "On Wisconsin!"

• POOMA - a high-performance mathematics C++ framework
• Blitz++ - a C++ mathematics library which uses template metaprogramming to achieve FORTRAN-caliber performance.
• MTL - another example of template metaprogramming.
• oonumerics.org - a good site for information on high-performance object-oriented code.

If the bios already has a basic serial driver

All BIOSes has had that as long as I remember. (Which is more than 10 years.)

why not a specialized tcp/ip stack, vnc, and ethernet driver as well?

They are quite limited in the amount of code it has space for. Originally they could only keep at most 64KB. Today some are manufactured with more, but it can only be used during boot before loading the OS. When the OS gets loaded it switches to the small size, so all drivers must fit into 64KB to be available to the OS.

This is also not an area in which much development is going on. In the days of DOS, the drivers in the BIOS was actually being used all the way. But today every OS has its own drivers, the BIOS is only being used during boot. So as soon as the computer can boot and get the OS into memory, nobody really cares about the BIOS any more. The BIOS API used by DOS and loaders has changed very little in the last years. This API does not include networking, and it probably never will. It has been many years since a new device got available through this API, most changes are just to deal with development in already supported units. Today BIOSes does support harddisks larger than 504MB. The last new device that has been added support for in the BIOS is the ATAPI CDROM, but this is really only suitable for boot, and it actually didn't change the API, it merely emulates a floppy.

Wasn't there a project to put linux in the bios along with a primitive firewall?

Yes.

http://www.acl.lanl.gov/linuxbios/news/index.html# packbot2

But I must be off now. I've got a YBa_2Cu_3O_{7-x} widget factory to get off the ground. :B

On a similar note, LinuxBIOS is some sort of ubercompressed, stripped-down linux kernel with some hardware-specific instructions that replaces a conventional BIOS, and then can activate a second-level boot loader like LILO. I don't know motherboard models, so I'm not sure if the project is geared more toward servers. The list of supported motherboards is here.

He started a group known as The Amateur Sky Survey, which has been working on software for analyzing the images from his cameras. After three years of scanning the celestial equator, we published a paper containing over 10 million measurements of stars in several passbands. You can read a preprint or the paper itself if you subscribe to PASP.

Based on our experience, I'd say that one of the hardest things about turning a backyard observatory into a serious scientific instrument is the bookkeeping: carefully recording all the necessary information and calibrating your results against the standard catalogs is a real pain, and doesn't have the same sex appeal as building the hardware or the software. But it's just as necessary.

We already fixed this problem when CompuServe freaked out about GIFs. We all use PNG now for everything.

Here's the problem. PNG is a good solution when you want lossless compression, which is (not coincidentally) exactly what GIF did. This matters when you actually care about the numbers in the image, or when you've got certain kinds of content (sharp edges and smooth color gradients in particular). JPEG is appropriate when you have pictures that look more like the real world... lots of variation, certain frequency distributions, tons of detail. The reason it works so well is that it removes information that you-the-user can't see and then does lossless compression on the rest. That is something that PNG doesn't do.

IANAL, but offhand I see two good bets for defeating this patent claim. The first is that it expires in two years: it was granted in 1987. The other is prior art: the original JPEG group was formed in 1985 by combining CCITT and ISO working groups trying to do roughly the same thing. (Source: The History of ISO 10918. I wasn't there; I'd appreciate corrections.) This patent seems to cover most of the components of JPEG and some of MPEG, and I just can't imagine that the JPEG committee hadn't come up with at least some of that by 1987.

That said, I do hack video and image encoders but I'm not a lawyer. I hope to see this claim shot down in flames. Quickly. I'm bothered by the idea that someone could out of the blue come and claim patent rights over my dissertation before I even finish it.

More info on the naming issue, and here.

yeah, I bet it'd look a lot like this, only faster.

It is sometimes assumed that the rapidity of biogenesis on Earth suggests that life is common in the Universe. Here we critically examine the assumptions inherent in this if-life-evolved-rapidly-life-must-be-common argument. We use the observational constraints on the rapidity of biogenesis on Earth to infer the probability of biogenesis on terrestrial planets with the same unknown probability of biogenesis as the Earth. We find that on such planets, older than ~ 1 Gyr, the probability of biogenesis is > 33% at the 95% confidence level. This quantifies an important term in the Drake Equation but does not necessarily mean that life is common in the Universe.

Take a look at http://sss.lanl.gov/pubs.shtml. The first paper there introduces the ToPPeR (TOtal Price-PErformance Ratio) Metric, and includes some bang-for-the-buck comparisons, primarily for other Beowulf clusters. The second one has some number comparing the performance-power and performance-space ratios of Green Destiny and several of the ASCI machines.

None of these have numbers for ASCI Q, however. But, please note that ASCI Q is only running at a couple of teraflops right now- it is not expected to be up to full speed for several months at earliest. On the other hand, Green Destiny took a couple of weeks to get up and running, as opposed to several years.

As I recall, poweron to Linux in single user mode was less than 5 seconds. That speed was largely a factor of how fast the code could be read from the EEPROMs.

Please ask Dell, HPQ, IBM, Gateway, and your favorite mother board manufacturers to dump the crappy old BIOSes and migrate to something modern.

There's an open source group doing it.

I dunno about commercial BIOS code, but the BIOS I use is pretty good... and I can verify it for myself if I want to. :-)

Hey, They're even bulding supercomputers with Transmeta processors now, like here .

Don't know if it runs linux though.

Check out www.linuxbios.org and you'll see that you DON'T need a BIOS to run your OS. Since modern OSes don't use the BIOS anyway, you can replace it with your bootloader and boot your machine in less than 3 seconds.

The big beowulfs all seem to be moving to this method now (combined with wake-on-lan) to save power.

Perhaps calling the other guy a moron was a bit, um, premature?

It's pretty funny to read that life isn't that common although other scientists claimed the opposite a couple of weeks ago. Here is another bit with some comments (sorry, german, use the fish).
Alex.

Space.com reports that life in the universe may be more rare than previously thought

I was recently reading up some more on gamma ray bursters, which are a recently discovered thing with explosions (so far only seen a very long way away) that appear to have an amount of energy equivalent to about the rest of the Universe put together.

There was a paper published in 1999 that theoriesed that every gamma ray burst was a galactic scale mass extinction event, and then attempted to extrapolate a rate that they occur locally in the Milky Way, then going on to suggest that because the rate is slowing down, we might be in a transition period for intelligent life appearing. It's all entirely theoretical, but it's an interesting read.

The good news is "at last we're here". The unfortunate bad news if the theory is correct is that because the last burst is somewhat overdue, we might not be here for much longer.

For what it's worth, there's a hugely massive star (eta carinae) about 7,000 to 10,000 light years away that's arguably ready to blow some time in the next million years. (If you're in the southern hemisphere it's a really nice thing to look at with binocs or better.) It's on the fringes of the theoretical limits of how massive a star can be, it's gone past the theoretical limits of the maximum amount of light that a star can possibly emit, and it's been suggested as a possible source of a future gamma ray burst in the Milky Way. Really though, nobody's quite sure what's about to happen. On the other hand we should probably be hoping that we're not nearby when it decides to go.

It's just another theory.

here and here.

Finally a changing speed of light is predicted in a DSR approach here.

here and here.

Finally a changing speed of light is predicted in a DSR approach here.

We chose the four clusters given here because they are representative of the production clusters that have been used at LANL over the past few years. The Acquisition costs were calculated based on the fact that the other clusters were all 'home-brew' purchases in pieces, leading to (at least initial) system admin costs to put the thing together and get it running. Furthermore, there's the inevitable burn-in period where some stuff fails and has to be replaced (costing sysadmin time). On the other hand, the RLX solution was sold as a whole and only had to be unpacked and plugged in, and other than one bad blade out of 240 (which was replaced within 24 hours) it's run flawlessly. Our experience purchasing pre-built clusters has been less than stellar and we've paid through the nose (thus the home-brew route).

The remaining sysadmin cost differences are because failures in the other solutions are expensive to fix. RLX hardware has hot swappable everything-- a blade dies, you plug another one in and it will automatically image the hard drive and come up with the appropriate hostname (based on rack/chassis/slot number) after a few minutes. After the initial setup, a monkey could administer this thing. The point is, this hardware was designed for failure cases, while the other solutions out there are still too close to the desktop paradigm. Oh, and one more thing-- we have a bladed solution by another company (18 tualitin 1.13Ghz blades in a 3U chassis) that sucks rocks. It runs so hot that running a piece of scientific code on for 10 minutes won't give you the correct answer with three runs. This technology is hard to get right, and while RLX may not have the perfect solution, they're on the right path IMNSHO.

in your favourite format.

in your favourite format.

in your favourite format.

I went and read their tco estimation in their whitepaper and came across something that really made me question their conclusions.

They compare tco for 24 node clusters of different architectures of beowulfs against the bladed cluster. The biggest expense by far for the traditonal systems is sysadmin time, over half, this after they spend most of the article talking about power. They estimate sysadmin costs for each of the traditional beowulfs at $60k over a 4 year period, while the bladed cluster at $800. Where does the $800 come from? They say that they haven't had to do any maintence on their system in the 9 months its been running! That doesn't sound like a very scientific data sampling to me.

There are other bladed designs, non-transmeta based, presumably the sysadmin costs would be the same. The last chart demonstrates that sysadmin costs are what's important, and that power, space, and downtime not nearly so.

Has anyone thought of doing this with Linux?

Looking at it this way, this system is simply a clearinghouse for problems people find interesting. If people work to break these big problems into manageable and concrete pieces, then these little pieces may be perfect for undergraduate or even graduate students to work on. I know that such a clearinghouse would be very valuable in mathematics. I've always imagined that such a thing would exist before too long for the mathematical community, and this would be a good thing.

The main reason that I think this would be a good thing is that for a young researcher starting out, one has to spend a lot of time understanding the big picture of a certian field, and generating good open problem on one's own. Such a system could bring the problems to the researchers more quickly. This could speed the process up by quite a bit. Such a thing sort of exists already, in the form of preprint servers, and I'm sure there's more to come! What doesn't really exist now is such a "big picture" fremework in a public domain, IMHO.

Take a look at this Cluster-in-a-lunchbox
aka BentoBox

They laughed at Einstein. They laughed at the Wright Brothers. But they
also laughed at Bozo the Clown.
-- Carl Sagan

Yes, most of the radical ideas in science have been ridiculed at one point in time. And scientists have regrettably supressed legitimate research because of their personal egos. A good (and unfortunate) relatively recent example is Brian Jospephson. It is unfortunate that this happens, and can really be hell on the people involved. Scientists are humans too, and sometime that gets in the way -- just like in every other field of human endevour.

But the thing in science is that if your theory is right, eventually people gather evidence for it, and scientists *do* accept it. And for every one misunderstood genius, there are hundreds of genuine crackpots who really get ridiculed more than they deserve, but still shouldn't be published alongside good research. Which is the whole purpose of peer review: to collect the small fraction of material that is interesting to the readers of the journal.

If you want to see what happens without peer review, go visit the lanl archives, which are a very valuable resource, but only if you realize that a large fraction of what is there is, while often written by respectable "Profs", wrong.

Propose another system that works better, and, if it does, it will evetually be adopted.

Including John Koch whom solved the century-old Four Color Problem

Ummm...the Four-Color Problem is not something one "solves". The "problem" was that the Four-Color Conjecture, while practically obvious to anyone, was very difficult to prove mathematically. And unless this guy changed his name, he isn't either of the dudes who proved the conjecture:

"In 1976, the conjecture was apparently proved by Wolfgang Haken and Kenneth Appel at the Univeristiy of Illinois..."

Sorry but los alamos is not in nevada - but NM. New Mexico

A bit of advice for trolling- change your name to something like:

Ray Juzaitis
Associate Laboratory Director
Weapons Physics
Los Alamos National Laboratory
Los Alamos, New Mexico
http://www.lanl.gov

He's a real guy, so it has to be believable.

One comment: even if you need to cool your detector to cryogenic temperatures, you don't have to have your customer pour liquid nitrogen (or did they say liquid helium?) into the commercial device. This is what compact no-maintenance closed-cycle coolers are for.

Plug #1: idQuantique
Plug #2: Magiq Technologies
Plug #3: Los Alamos lab (yes there used to be a site there)
Plug #4: Our own research (not commercially-oriented yet)

I'm partial to this version myself.

Quantum cryptography at other places:

Quantum Information at Los Alamos National Laboratory

Yup. "NNSA HQ has requested that LANL review all publicly accessible information".

Acer, SiS and VIA have been supportive with both hardware and software for the LinuxBIOS project. From what I remember reading on the mailing list a while back, the engineers were quite helpful.

You want to see what Gabe Luther and and William Towler,current "holders" of Big G, used to measure it ? Here's a great shot of the torsion balance device from this short summary .

Here's a link to the press room at LANL Look for "17) Measuring the Gravitational Constant ("Big G") -- In the Lab of Gabe Luther, Los Alamos scientist. Sound bite on methodology." - no link but an interesting page of resources.

The easiest way to demonstrate this is to note that NASA has invested in research [space.com] to try to replicate Podkletnov's [amasci.com] results.

And have been uncessful. Further NASA apparently gave support to Puthoff (the guy who proclaimed Uri Geller legit). His ``paper'' can be found here (download the pdf and check the acknowledgements). That doesn't exactly inspire confidence in NASA's jusdgement.

The interesting thing about gravity is that it isn't well understood by modern physics.

Totally untrue. General relativity is understood very well. Predictions for Binary pulsar systems agree very well with observation (many decimal places).

We know how it behaves (we think) but we don't know what causes it really.

The curvature of spacetime. We've known that since 1915.

This makes it equally ripe for psuedo-science as for breakthrough science.

Any breakthroughs are going to come in an extreme regime, much like the deviations from newtonian physics. You're not going to see macroscopic violations of GR.

Matthew Nobes (website)

http://www.lanl.gov/worldview/news/releases/archiv e/02-028.shtml

In particular, the lack of pulsation isn't quite the only thing pointing to something odd going on (as I'm sure you're aware, but some people might not be). They find a fairly good spectral fit to a 60 ev (700,000-ish K) blackbody, which yields a radius less than the 10-12 km or so allowed by current NS theories; while I haven't really gone over their paper in detail, they claim to have ruled out two-component blackbodies, at least at any level that would contribute appreciably to the flux, and power-law sources at high confidence. And while there remains some question as to the distance (the Walter (2001) measurements of 60 pc versus Kaplan et al (2002)'s 140 pc or so), I think their arguments in support of the larger distance (e.g., the larger distance is more in agreement with neutral Hydrogen column measurements plus standard physical density estimates) are reasonably compelling, albeit prone to criticism.

I'd be curious to hear your thoughts on this -- i.e., do you think a much-lower temperature blackbody (or "hot spot" model) is not truly excluded by the data on this line of sight? Because the lack of pulsation here is just one part of the puzzle. :-)

Cheers.

For those of you with the stomach for it, here's the preprint.