I can't be the only/.er who remembers programs that printed Snoopy Calendars on text-only line printers by exploiting over-printing and the different densities of the various characters.
So, it is a real question: could one find settings for the Difference engine + printer that would print a Snoopy Calendar? or any recognisable image?
Is there a simulator available?
Actually that's an ever better question, and a nice student project: write a Difference engine (or even Analytical Engine) simulator -- ideally with graphics showing the simulated machine in operation, but failing that, just simulating the abstract operations.
Back in the 80s I had an (unclassified) summer job for a UK government agency which did a lot of secure stuff. One of their core security policies was the "high water mark" policy, which said that a physical document could only go up in classification, never down (although it might be possible to get permission to make a less classified document with the same contents). For paper documents this is really sensible: it is NEVER acceptable to have a document with SECRET crossed out and CONFIDENTIAL written in, for instance. The problem came when they (a) applied this to magnetic media and (b) treated entire physical (hard) disks as single documents. This meant that if a disk had ever had 1 byte of TOP SECRET information on it, then the whole disk, forever, was TOP SECRET. The inevitable result was that disks slowly migrated up the security levels, and there was always a glut of TOP SECRET disks and a shortage of unclassified ones. The surplus TS disks were eventually taken out, hit with a hammer a bit and chained to the wall at the back of the computer building, as no one had yet approved a method of disposing of them.
Both these problems are really economic, not technological, in the sense that solutions are out there, they just cost a lot more than the big disks.
If you look at expensive server systems, let alone at mainframes, they already have solutions for these problems -- for I/O you do pretty well with U160W SCSI and 66MHz/64bit PCI; for backups, you put your storage and your backup device on a SAN, for which Gigabit ethernet is pretty much entry-level, and your backup device is a tape jukebox (or you just mirror your disks heavily and forget conventional tape backups).
The anomaly is that top-end disk technology has come out very cheap, thanks I guess to the huge volumes that are shipping, and so you have 2000 dollar PCs with disks that really "belong" in 10000 dollar servers.
Re:COmpetition? There is no competition!
on
1.4-1.6 GHz Alphas
·
· Score: 1
Look at http://www.specbench.org. On SpecCPU95 (there aren't enough SpecCPU2000 results to be worth worrying about), top Intel and AMD processors lead on integer performance, followed by Alpha, with HP a poor third and Sun nowhere.
On FP performance, though it's a different story. There Alpha lead, followed by HP, then AMD, Intel and Sun.
Umm. 99.999% uptime is 1 in 100000 downtime. This is 0.86 seconds per DAY, or more reasonably, about 5 minutes per year. This is imaginable for a permanently manned installation. Once every year or so, you could toleate the the operator having to manually switch off a malfunctioning CPU that had refused to failover gracefully, or remove a dodgy cable that was causing behaviour too unpredictable for the O/S to figure it out, or something of that order.
No one seems to have mentioned the probable direct detection of reflected light from the planet around tau Bootis. This is reported in Nature, vol 402, 16 December 1999. The main author's office is just across the corridor from me.
This was mainly, a monumental data processing achievement. They described it on a recent TV program as "like detecting a firefly sitting on the edge of a searchlight 10 miles away". As I understand it, they modeled the way that the reflected light would change with both orbital geometry and Doppler shift and used very sophisticated processing to extract a signal with those characteristics from the background time-resolved spectrum of the star.
One bonus of this work is that the planet's signal is much stronger in green light than in red or blue, giving an idea of its colour. The same approach might eventually allow for detection of the presence or absence of specific compounds on the planets surface.
The authors don't see this approach generalizing to find Earth-like planets. They are too dim, and their orbital velocities are low enough that the Doppler shift is lost in the shifts due to movement of gas on the Sun's surface. On the other hand, they thought that interferometric work in the medium IR stood a good chance. You would need a flottile of perhaps 4 8m space telescopes, with supercooled detectors and extremely accurate station-keeping (to with 1 micron over extended periods) but you could then use interference to cancel the image of the star almost completely, while preserving any light from the planet.
The Open University, although based around rather older technology is an interesting model. It was set up in the 1970s and uses, TV, radio (and these days distribution of video and audio tapes) and printed course notes to deliver a pretty full range of courses, leading to a British degree after (normally) 6 years of part-time study.
Their qualifications are well-respected, and their audio, video andprinted course material is top quality, but it is not a cheap thing to do. Firstly the production costs of good material are high -- they employ top lecturers and professional TV production crews and a much higher level of display equipment and technical support than you'd see in a "face-to-face" lecture, and the course material has to be revised every 5-10 years depending on subject -- if nothing else, the lecturer's clothes and hair styles start to look rather laughable.
They also employ tutors all over the country to run face-to-face small group tutorials (I think most students get about 2 per month) or, where people are very scattered, audio, video or internet chat conferences.
You can make distance learning work really well, and the internet helps a little, but it is not dramatically less work, or cheaper, than a conventional university.
this law needs to be passed at the federal level, so that it affects all 50 states
I hate to mention it, but there is an Internet outside the USA. Would the federal version of this law require someone in (say) Liberia sending email to an address which turns out to be in Uzbechistan to check whether it's actually in the USA? If so, how is this different from the problem with the State law.
To regulate the Internet effectively, especially the non-commercial aspects, would need a world government, or a complex multi-lateral treaty and an international enforcement agency. The former seems unlikely in the near future; the latter would be too slow to set up and too slow to change. Perhaps if the Internet stabilises.
Some regulation is possible in the commercial level. The UK, for instance, could legislate to require UK credit card companies to extract "no spam" contracts from the merchants they deal with, making it hard for spammers to collect money for whatever they're trying to sell. On the other hand, if you make it too expensive to be a UK crredit card company, offshore operations would just set up and take their business.
You have a fair point. The 1985 media could have been copied in 1995 onto newer media a fraction the size. If this was not done then a problem arises. It's not insuperable, unless the media decay, but it can be expensive. We can always hand-build a tape drive, or read the magnetic field directly off the tape with an electron microscope, but it costs.
Books are not as good as you think though. People needed to (a) be able to read and (b) speak the relevant language, for archives of old books to be of any use. Neither is completely automatic. Also, to make real use of old books, the readers would need a fair amount of cultural context for them, and that is positively expensive to acquire.
At the moment, Moore's law is the only thing that stops this problem becoming really acute. Although I keep all my email, and the total size of the archive grows almost exponentially, so does the size of my hard disk, and the speed at which I can run grep over it.
To handle terabyte databases now, needs leading-edge hardware and state-of-the-art software specially optimised for the data format. In 20 years, however, we will just be able to haul the terabyte database into emacs, and hack up some macros to reformat it and search it.
If Moore's law ever tops out, then we are in trouble!
A year or two back, I set an exam question in which I gave some measurements and asked what aspect of a PC they were likely to apply to. One was 200MHz. One student claimed this was the rotation speed of the floppy disk. At this speed, the edge of the disk is doing about 3% of light-speed.
The manufacturers FLY the drives to the US using cargo 747s
Has anyone ever worked out the bandwidth of one of these planes if you loaded up the drives with data before shipping. I guess the drives weigh 100g each, and a cargo 747 must be able to carry 100 tons, so thats approx 1 million drives (reality check, this is a pile 10m x 15m x 2m, should fit in easily), so a single flight carries about 18 PB (peta bytes, 10^15 bytes) of data. Say it can make one delivery every other day, that is about every 180000 seconds, so we get a bandwidth of 100GB/s, or 800 Gb/s -- who needs project Oxygen (320 Gb/s transocean cabling) anyway?
"Latency" do I hear someone ask? "Don't be small minded!" I say. Latency can be dealt with by proper caching strategies at a higher level of the protocol stack.
Steve Linton
PS a 100 000 ton cargo ship full of these things does even better.
Propane is a liquid at room temperature and quite modest pressure, which is undoubtedly how it is stored in a cars fuel tank. If the pressure escapes, a proportion will boil off, cooling the rest until it is liquid at atmospheric pressure (about -40C if I recall correctly) and it will then sit there boiling off relatigely slowly.
Hydrogen is not so obliging. It will not liquefy under any (sane) pressure at room temperature, and will boil very quickly (low latent and specific heats) if the cold liquid gets warm.
Something like a scuba tank is one of my options, but having one in a car does present some weight and safety questions.
One option is to store it as a very cold liquid, requiring a well insulated fuel tank. This is possible, but a bit hard to handle safely. If the tank cracks, or some heat gets in, you will have rather a lot of hydrogen gas trying to get out, and hydrogen gas is explosive in proportions of something like 4% to 80% in air.
Another is to store it under moderate pressure (a few atmospheres) adsorbed onto the surface of a metal dust. In this model your fuel tank is lightly pressurized, and full of dust. You pump hydrogen gas in under pressure, and it is adsorbed, but when you let some out and reduce the pressure slightly, it is released again. The problem here is that the tank is heavy.
The third option is to store it as a gas, under very high pressure. This requires a really serious pressure vessel as your fuel tank, which is likely to be heavy, and you will need to engineer the tank to seal itself and remain intact in a crash, adding still more weight, and a rather heavy object flying through the wreckage flattening people.
It might be better to use the hydrogen to make something which is a bit easier to store and not too much more polluting, like methanol.
The light we see is probably in the red or near-IR, but it will have been emitted as far UV light. In the meantime it has been red-shifted all the way down.
The much talked about Cosmic Microwave Background started out as UV, when hydrogen atoms deionized a few thousand (? might be a few million, not sure) years after the Big Bang but has now red-shifted all the way to microwaves
1) If huge bales of pure gold metal were floating in orbit and all you had to do was open the shuttle doors and scoop it up with rakes, it would still come nowhere near offsetting the cost of a shuttle launch. Or the next generation of "cheap" vehicle launches. Moving mass off earth or back to Earth is INSANELY EXPENSIVE.
No one is seriously proposing such a process. In the short term mining would be for water (probably in the form of water ice) for use in space (electrolyzed, as fuel and oxygen, or "straight" to drink).
Moving mass off Earth will stay expensibe for a while, but getting it back could become cheap rather quickly. You only need to drop something slightly below orbital velocity and it will start being slowed down by the atmosphere. Simple controls will crashland it on the desert or ocean of your choice. Not much more is needed to soft land.
Many/most asteroids are made of fluff, crap, and dust. They are not rocks. They are not mine-able, and they do not offer a place to land.
All the better. For most purposes you actually want a mini-comet, say 10m across and mostly made of "dirty snow". A mining probe simply floats near it, puts a "bag" over all or part of the surface and starts heating it up with reflected sunlight. Then pipe the steam round into the shadow of the mirror and condense it. Rock is juts a nuisance, and landing is just a way to cope with too much gravity.
Re:Why not just use the Crusoe as a G4?
on
Darwin on Crusoe?
·
· Score: 1
There has been a lot of hype about getting the Crusoe to emulate non x86 architectures, but I am actually quite suspicious of whether this would work well. There are some features of the Crusoe hardware, like the set of condition flags and the memory management that are actually the same as the x86 equivalent, because emulating those things was too expensive.
Something like a native JVM implementation is reasonable, because that is at a much higher level, but I think emulating non-intel processor instruction sets is very unlikely.
The basic reason why compilers are slower is a programmer can use information that is simply not in the program in their optimization.
In one (in)famous example, a quick, but, in general unsafe conversion was used in part of the Ariane 4 control software, justified by a mathematical formula present in a comment which proved from the geometry of the spacecraft that it was, in fact always safe. [ sadly, Ariane 5 had different geometry.....]
A more down to Earth example is a loop that you may know will never execute 0 times, which would allow you some optimisation.
Profiling JIT compilers can do somewhat better, in that they can, at least, note that condition X has never happened to date, and produce code that handles it as an expensive exception, while concentrating on keeping the main path fast. The Transmeta stuff seems to do a lot of that.
If the tech you're trying to patent has been used before, you won't get your patent.
This is not so. The tech has to have been published. This goes back to the original purpose of the patent system, which was not so much to reward innovation as to encourage its sharing. If you had an unpublished technique and were prepared to publish it, then you got a 17 year monopoly on its use.
It will surely be better to get a MB now that works well for your Celerons (66MHz FSB, PC66 SDRAM) and save any money you can. then dump the lot and buy a new MB and memory when you buy your new processors. By then you will want 133 or 166 MHz FSB (or something) or you might want to go the Athlon route, and who knows what memory will do.
Get what you need now and plan on replacing it all -- you'll get a better system for little more money
The thing that matters for losses is voltage. The higher the voltage the less loss. AC won out early on because it is easier to convert AC to higher voltages and back again.
Today DC is starting to come back in because we now have technology to change voltages efficiently for DC, and very high voltage AC has a problem with arcing that is less bad for DC. A 1MV DC line has a peak of 1MV while a 1MV AC line has a peak of 1.44MV and so can arc further.
Radiation losses are not an issue for AC power because the frequency (50 or 60 Hz) is so low. It is an issue for high-speed data transmission over power lines.
A paper is being refereed for Nature. The BBC actually broke an embargo on the news (as the Sunday Times did a few weeks ago).
I've been to a few talks by the guy who did the work (his office is just across the corridor from mine) and their basic expertise is in very detailed analysis of time-variations in the exact shape of spectral lines. A spectral line from a star is not actually a line. It gets broadened by Doppler shift where the atoms emitting (or reflecting or absorbing) the light are moving towards, or away from, the observer.
To take an example from another project of theirs, consider a rapidly rotating star with a dark "sunspot" at its equator. As the spot comes into view around the side of the star, the blue-shifted part of the spectral line will get a little dimmer, then, as the star rotates, the "dimming" will shift to the centre of the line, and then towards the red-shifted side.
What they do is to use a very precise spectrometer to record the distribution of light across the line at regular intervals. This gives them a big two-dimensional data set (dimensions are frequency and time). They then crunch this with some fairly serious computational power (I asked Prof Cameron why they had got in first on this, and he said that they were the only group with the software) to extract a best guess at the distribution of spots on the star. Phrases like "least squares deconvolution" and "maximum entropy" often appear on his noticeboard.
The same technique applies to detecting the light from the planet. Such light would move through the spectrum in a predictable way, and they could use their software to search for it. The fact that they saw the same pattern in several spectral lines is a good confirmation of the real planet.
Crucial to success in this case was (a) a fairly bright star (b) a big planet (c) a planet moving fast in its orbit.
To digress a little, the increasing number of extrasolar planets being discovered is, in a way, encouraging to those who hope that we are not alone, but also troubling. The planets being detected at present are (a) big (b) close to the star (c) often in quite eccentric orbits. Earth-like planets might have real problems co-existing with these monsters. Also, I heard a suggestion that magentic and tidal effects from big close planets may cause occasional "hyper-flares" which would sterilise Earth-like planets quite effectively.
Slashdot Mirror
I can't be the only /.er who remembers programs that printed Snoopy Calendars on text-only line printers by exploiting over-printing and the different densities of the various characters.
So, it is a real question: could one find settings for the Difference engine + printer that would print a Snoopy Calendar? or any recognisable image?
Is there a simulator available?
Actually that's an ever better question, and a nice student project: write a Difference engine (or even Analytical Engine) simulator -- ideally with graphics showing the simulated machine in operation, but failing that, just simulating the abstract operations.
Steve Linton
Back in the 80s I had an (unclassified) summer job for a UK government agency which did a lot of secure stuff. One of their core security policies was the "high water mark" policy, which said that a physical document could only go up in classification, never down (although it might be possible to get permission to make a less classified document with the same contents). For paper documents this is really sensible: it is NEVER acceptable to have a document with SECRET crossed out and CONFIDENTIAL written in, for instance. The problem came when they (a) applied this to magnetic media and (b) treated entire physical (hard) disks as single documents. This meant that if a disk had ever had 1 byte of TOP SECRET information on it, then the whole disk, forever, was TOP SECRET. The inevitable result was that disks slowly migrated up the security levels, and there was always a glut of TOP SECRET disks and a shortage of unclassified ones. The surplus TS disks were eventually taken out, hit with a hammer a bit and chained to the wall at the back of the computer building, as no one had yet approved a method of disposing of them.
Both these problems are really economic, not technological, in the sense that solutions are out there, they just cost a lot more than the big disks.
If you look at expensive server systems, let alone at mainframes, they already have solutions for these problems -- for I/O you do pretty well with U160W SCSI and 66MHz/64bit PCI; for backups, you put your storage and your backup device on a SAN, for which Gigabit ethernet is pretty much entry-level, and your backup device is a tape jukebox (or you just mirror your disks heavily and forget conventional tape backups).
The anomaly is that top-end disk technology has come out very cheap, thanks I guess to the huge volumes that are shipping, and so you have 2000 dollar PCs with disks that really "belong" in 10000 dollar servers.
Look at http://www.specbench.org. On SpecCPU95 (there aren't enough SpecCPU2000 results to be worth worrying about), top Intel and AMD processors lead on integer performance, followed by Alpha, with HP a poor third and Sun nowhere.
On FP performance, though it's a different story. There Alpha lead, followed by HP, then AMD, Intel and Sun.
Horses for courses.
Umm. 99.999% uptime is 1 in 100000 downtime. This is 0.86 seconds per DAY, or more reasonably, about 5 minutes per year. This is imaginable for a permanently manned installation. Once every year or so, you could toleate the the operator having to manually switch off a malfunctioning CPU that had refused to failover gracefully, or remove a dodgy cable that was causing behaviour too unpredictable for the O/S to figure it out, or something of that order.
Still pretty impressive though.
Eclipse detection is great if the planet's orbit
is sufficiently well aligned with the line from the star to Earth that we get eclipses.
Unfortuately, this gets less and less likely, the smaller the planet is, and the further from the star.
This was mainly, a monumental data processing achievement. They described it on a recent TV program as "like detecting a firefly sitting on the edge of a searchlight 10 miles away". As I understand it, they modeled the way that the reflected light would change with both orbital geometry and Doppler shift and used very sophisticated processing to extract a signal with those characteristics from the background time-resolved spectrum of the star.
One bonus of this work is that the planet's signal is much stronger in green light than in red or blue, giving an idea of its colour. The same approach might eventually allow for detection of the presence or absence of specific compounds on the planets surface.
The authors don't see this approach generalizing to find Earth-like planets. They are too dim, and their orbital velocities are low enough that the Doppler shift is lost in the shifts due to movement of gas on the Sun's surface. On the other hand, they thought that interferometric work in the medium IR stood a good chance. You would need a flottile of perhaps 4 8m space telescopes, with supercooled detectors and extremely accurate station-keeping (to with 1 micron over extended periods) but you could then use interference to cancel the image of the star almost completely, while preserving any light from the planet.
Surely this doesn't let Yahoo off the hook, it just lets Yahoo sue the sellers for whatever they have to pay Nintendo.
Steve
Their qualifications are well-respected, and their audio, video andprinted course material is top quality, but it is not a cheap thing to do. Firstly the production costs of good material are high -- they employ top lecturers and professional TV production crews and a much higher level of display equipment and technical support than you'd see in a "face-to-face" lecture, and the course material has to be revised every 5-10 years depending on subject -- if nothing else, the lecturer's clothes and hair styles start to look rather laughable.
They also employ tutors all over the country to run face-to-face small group tutorials (I think most students get about 2 per month) or, where people are very scattered, audio, video or internet chat conferences.
You can make distance learning work really well, and the internet helps a little, but it is not dramatically less work, or cheaper, than a conventional university.
To regulate the Internet effectively, especially the non-commercial aspects, would need a world government, or a complex multi-lateral treaty and an international enforcement agency. The former seems unlikely in the near future; the latter would be too slow to set up and too slow to change. Perhaps if the Internet stabilises.
Some regulation is possible in the commercial level. The UK, for instance, could legislate to require UK credit card companies to extract "no spam" contracts from the merchants they deal with, making it hard for spammers to collect money for whatever they're trying to sell. On the other hand, if you make it too expensive to be a UK crredit card company, offshore operations would just set up and take their business.
You have a fair point. The 1985 media could have been copied in 1995 onto newer media a fraction the size. If this was not done then a problem arises. It's not insuperable, unless the media decay, but it can be expensive. We can always hand-build a tape drive, or read the magnetic field directly off the tape with an electron microscope, but it costs.
Books are not as good as you think though. People needed to (a) be able to read and (b) speak the relevant language, for archives of old books to be of any use. Neither is completely automatic. Also, to make real use of old books, the readers would need a fair amount of cultural context for them, and that is positively expensive to acquire.
At the moment, Moore's law is the only thing that stops this problem becoming really acute. Although I keep all my email, and the total size of the archive grows almost exponentially, so does the size of my hard disk, and the speed at which I can run grep over it.
To handle terabyte databases now, needs leading-edge hardware and state-of-the-art software specially optimised for the data format. In 20 years, however, we will just be able to haul the terabyte database into emacs, and hack up some macros to reformat it and search it.
If Moore's law ever tops out, then we are in trouble!
I am pretty sure. I might have the numbers slightly wrong, but it's a wide range. Hydrogen is WORSE than acetylene in this regard.
A year or two back, I set an exam question in which I gave some measurements and asked what aspect of a PC they were likely to apply to. One was 200MHz. One student claimed this was the rotation speed of the floppy disk. At this speed, the edge of the disk is doing about 3% of light-speed.
Has anyone ever worked out the bandwidth of one of these planes if you loaded up the drives with data before shipping. I guess the drives weigh 100g each, and a cargo 747 must be able to carry 100 tons, so thats approx 1 million drives (reality check, this is a pile 10m x 15m x 2m, should fit in easily), so a single flight carries about 18 PB (peta bytes, 10^15 bytes) of data. Say it can make one delivery every other day, that is about every 180000 seconds, so we get a bandwidth of 100GB/s, or 800 Gb/s -- who needs project Oxygen (320 Gb/s transocean cabling) anyway?
"Latency" do I hear someone ask? "Don't be small minded!" I say. Latency can be dealt with by proper caching strategies at a higher level of the protocol stack.
Steve Linton
PS a 100 000 ton cargo ship full of these things does even better.
Propane is a liquid at room temperature and quite modest pressure, which is undoubtedly how it is stored in a cars fuel tank. If the pressure escapes, a proportion will boil off, cooling the rest until it is liquid at atmospheric pressure (about -40C if I recall correctly) and it will then sit there boiling off relatigely slowly.
Hydrogen is not so obliging. It will not liquefy under any (sane) pressure at room temperature, and will boil very quickly (low latent and specific heats) if the cold liquid gets warm.
Something like a scuba tank is one of my options, but having one in a car does present some weight and safety questions.
One option is to store it as a very cold liquid, requiring a well insulated fuel tank. This is possible, but a bit hard to handle safely. If the tank cracks, or some heat gets in, you will have rather a lot of hydrogen gas trying to get out, and hydrogen gas is explosive in proportions of something like 4% to 80% in air.
Another is to store it under moderate pressure (a few atmospheres) adsorbed onto the surface of a metal dust. In this model your fuel tank is lightly pressurized, and full of dust. You pump hydrogen gas in under pressure, and it is adsorbed, but when you let some out and reduce the pressure slightly, it is released again. The problem here is that the tank is heavy.
The third option is to store it as a gas, under very high pressure. This requires a really serious pressure vessel as your fuel tank, which is likely to be heavy, and you will need to engineer the tank to seal itself and remain intact in a crash, adding still more weight, and a rather heavy object flying through the wreckage flattening people.
It might be better to use the hydrogen to make something which is a bit easier to store and not too much more polluting, like methanol.
The light we see is probably in the red or near-IR, but it will have been emitted as far UV light. In the meantime it has been red-shifted all the way down.
The much talked about Cosmic Microwave Background started out as UV, when hydrogen atoms deionized a few thousand (? might be a few million, not sure) years after the Big Bang but has now red-shifted all the way to microwaves
Moving mass off Earth will stay expensibe for a while, but getting it back could become cheap rather quickly. You only need to drop something slightly below orbital velocity and it will start being slowed down by the atmosphere. Simple controls will crashland it on the desert or ocean of your choice. Not much more is needed to soft land.
All the better. For most purposes you actually want a mini-comet, say 10m across and mostly made of "dirty snow". A mining probe simply floats near it, puts a "bag" over all or part of the surface and starts heating it up with reflected sunlight. Then pipe the steam round into the shadow of the mirror and condense it. Rock is juts a nuisance, and landing is just a way to cope with too much gravity.There has been a lot of hype about getting the Crusoe to emulate non x86 architectures, but I am actually quite suspicious of whether this would work well. There are some features of the Crusoe hardware, like the set of condition flags and the memory management that are actually the same as the x86 equivalent, because emulating those things was too expensive.
Something like a native JVM implementation is reasonable, because that is at a much higher level, but I think emulating non-intel processor instruction sets is very unlikely.
The basic reason why compilers are slower is a programmer can use information that is simply not in the program in their optimization.
In one (in)famous example, a quick, but, in general unsafe conversion was used in part of the Ariane 4 control software, justified by a mathematical formula present in a comment which proved from the geometry of the spacecraft that it was, in fact always safe. [ sadly, Ariane 5 had different geometry.....]
A more down to Earth example is a loop that you may know will never execute 0 times, which would allow you some optimisation.
Profiling JIT compilers can do somewhat better, in that they can, at least, note that condition X has never happened to date, and produce code that handles it as an expensive exception, while concentrating on keeping the main path fast. The Transmeta stuff seems to do a lot of that.
It will surely be better to get a MB now that works well for your Celerons (66MHz FSB, PC66 SDRAM) and save any money you can. then dump the lot and buy a new MB and memory when you buy your new processors. By then you will want 133 or 166 MHz FSB (or something) or you might want to go the Athlon route, and who knows what memory will do.
Get what you need now and plan on replacing it all -- you'll get a better system for little more money
Several confusions here.
The thing that matters for losses is voltage. The higher the voltage the less loss. AC won out early on because it is easier to convert AC to higher voltages and back again.
Today DC is starting to come back in because we now have technology to change voltages efficiently for DC, and very high voltage AC has a problem with arcing that is less bad for DC. A 1MV DC line
has a peak of 1MV while a 1MV AC line has a peak of 1.44MV and so can arc further.
Radiation losses are not an issue for AC power because the frequency (50 or 60 Hz) is so low. It is an issue for high-speed data transmission over power lines.
A paper is being refereed for Nature. The BBC actually broke an embargo on the news (as the Sunday Times did a few weeks ago).
I've been to a few talks by the guy who did the work (his office is just across the corridor from mine) and their basic expertise is in very detailed analysis of time-variations in the exact shape of spectral lines. A spectral line from a star is not actually a line. It gets broadened by Doppler shift where the atoms emitting (or reflecting or absorbing) the light are moving towards, or away from, the observer.
To take an example from another project of theirs, consider a rapidly rotating star with a dark "sunspot" at its equator. As the spot comes into view around the side of the star, the blue-shifted part of the spectral line will get a little dimmer, then, as the star rotates, the "dimming" will shift to the centre of the line, and then towards the red-shifted side.
What they do is to use a very precise spectrometer to record the distribution of light across the line at regular intervals. This gives them a big two-dimensional data set (dimensions are frequency and time). They then crunch this with some fairly serious computational power (I asked Prof Cameron why they had got in first on this, and he said that they were the only group with the software) to extract a best guess at the distribution of spots on the star. Phrases like "least squares deconvolution" and "maximum entropy" often appear on his noticeboard.
The same technique applies to detecting the light from the planet. Such light would move through the spectrum in a predictable way, and they could use their software to search for it. The fact that they saw the same pattern in several spectral lines is a good confirmation of the real planet.
Crucial to success in this case was (a) a fairly bright star (b) a big planet (c) a planet moving fast in its orbit.
To digress a little, the increasing number of extrasolar planets being discovered is, in a way, encouraging to those who hope that we are not alone, but also troubling. The planets being detected at present are (a) big (b) close to the star (c) often in quite eccentric orbits. Earth-like planets might have real problems co-existing with these monsters. Also, I heard a suggestion that magentic and tidal effects from big close planets may cause occasional "hyper-flares" which would sterilise Earth-like planets quite effectively.
Steve Linton