The main memory held 1000 words, but that was composed of 100 mercury channels. (The whole memory system had 126 channels, but some of the channels were not part of main memory.) Each channel was a single acoustic path through a mercury tank, with 18 channels per tank. A main memory channel stored 10 words, with each word composed of 12 characters (11 digits plus sign), each of which was 6 bits plus a parity bit. That comes out to 840 bits.
The note estimates 1000 total bits per channel based on the clock rate and the maximum access time, but I think he got his numbers a little wrong. (He used a clock rate of 2.5 MHz, but it should be 2.25 MHz, which gives an answer of 900 bits.)
To make the story short, Ken Thompson and Dennis Ritchie wrote the first PDP/10 and PDP/11 implementations of what was to become Unix in the early 70's, at Bell laboratories.
You're half right there. The first implementation of Unix was in 1969 on a PDP-7, not a PDP-10. The two machines are quite different: the PDP-7 was a small 18-bit minicomputer designed for small groups of users, while the PDP-10 was a large 36-bit mainframe-type machine designed to serve big communities, like an entire university.
It was natural for Thompson and Ritchie to move from the PDP-7 to the 16-bit PDP-11, since the two machines were of similar scale in the beginning. Moving from the PDP-10 to the PDP-11 would have been like moving from your nice roomy house into your college friend's attic.:-)
There were a number of interesting operating systems on the PDP-10 (TENEX, TOPS-10, TOPS-20), but no one has ever implemented a Unix-style OS on it, as far as I know.
Dennis Ritchie's web site has a good account of the early development of Unix. It makes interesting reading if you're curious about such things.
Quite a bit prior to that--it was last produced in 1959.
Unfortunately, the joke of Ford Prefect's name was lost on most Americans, myself included. I didn't even know the car existed until I read the obituary for Douglas Adams, and my dad (an Englishman old enough to remember the Prefect) mentioned the car.
Wikipedia has a pretty good entry covering both the car and Adams' choice of the name for his character.
Just to clarify, SPEC's CPU benchmarks aren't synthetic benchmarks. A synthetic benchmark is a program written to test performance that doesn't do any real useful work (for example, Dhrystone). SPEC's CPU benchmarks are real applications performing real application workloads (for example, running a particle accelerator simulation, or executing Perl scripts), so they actually provide some indication of how fast a computer system with a certain compiler can perform those kind of tasks.
The biggest problems with SPEC's CPU benchmarks is that they tend to concentrate on technical applications and that people only talk about the average SPECint and SPECfp scores, neglecting the individual benchmark scores that correspond to real tasks. But you can always find the individual benchmark scores on SPEC's website.
Wow, 108 comments and I have yet to find one that discusses the proposition in any detail. (maybe all those fluid dynamics equations are as foreign to other slashdotters as they are to me:-)
Yeah, I suppose the body of the paper is really aimed at other geophysicists. Sorry about that. I didn't try to thoroughly comprehend all his calculations either, since I've left the field and haven't studied any of this stuff in about a decade. But I think I got the gist of his ideas. Let me try to answer your points.
1) The reason you use all that iron isn't to keep the blob hot. (Actually, it stays hot by itself; as it falls, the gravitational energy released from its fall heats it up. But this extra heat is conducted away to the surrounding rock, so the blob won't get too hot and melt the probe.) To reach the core, the iron is pushing on the bottom edge of a vertical crack, forcing it open all the way down. But to force the crack open, you need a lot of weight. Hence the big mass of iron. A blob of radioactive material would have to be almost as big to do the job, and would be a lot more expensive and...hard to work with.:-)
Reading that part of the paper again, I notice that he's referring to an existing, accepted theory used for understanding the movement of molten magma upward through cracks in the crust. It works in the same way, only upward (because the molten magma is lighter than the surrounding rock, unlike the iron here, which is heavier). The point is that this part of the idea is based on an understood phenomenon.
2) I think that comment was meant to be tongue-in-cheek, just like the paper title.
3) Well, the crack closes up, but there's still going to be a narrow fracture where the crack was. The crack isn't going to instantly heal, since mantle rock at that depth and pressure is solid. And there might also be a small amount of iron left behind in the crack. As he said in his annotation, "This may be a problem with the whole idea."
4) I think the idea here is mostly to investigate the outer core, not the intervening mantle. (As I recall, the mantle is much better known than the core.) So he doesn't care about measurements until the probe reaches the core. Once the liquid blob reaches the liquid outer core, it will presumeably spread out and mix somehow, while the probe, being solid, will keep sinking for a bit, long enough to function for a while with appropriate engineering.:-) This is where things get fuzzy for me; since the probe was built to be neutrally buoyant in the blob, how does it cease to float around in blob material once it gets to the core? Would its material have to be chosen to change density at the proper pressure so that it would start to sink?
5) Actually, he's really saying that we don't know much about how the intervening material distorts a seismic signal. But the neat thing about digital signals is that as long as you can read the bits, you don't care too much about the shape of the waveform. That's not true when "imaging" earth's interior using the travel times of earthquake seismic waves. Also, the idea is to send the data using compressional waves, which travel through liquid just fine. Of course, you're right that you'd have to get lots of practice transmitting data seismically first.
Naturally, I realize this idea is heavily on the loony side. That's what made it fun to read and think about.:-)
Yes, magnetic drums were a very common storage device until disks took over in the 1960s and 1970s. By the way, drums generally had a line of fixed heads all the way along the drum, so there was no need to move the heads and thus no seek latency, only rotational latency. And while drums were as big and noisy as washing machines, so were the disks of that era.
Magnetic drums weren't used for quite the same purposes as disks, though. Disks were for file storage, but drums were more often used as a low speed high-density working memory, similar to a modern virtual memory swap device. Running programs were swapped between core memory and the drum.
Before magnetic core memory became cheap enough to use in low-cost computers, some people even built low-end computer systems with a magnetic drum as the only working memory. Such computers were much slower than computers with electronic memory, since you might have to wait an entire rotational delay for a desired memory word to come under the read heads. Clever programmers arranged the instructions and data on the drum memory to minimize this delay.
To me, it was quite reasonable for the British government to keep Colossus a secret, and it didn't cost Britain much.
Remember, Colossus was developed as a code-breaking enterprise, not as a computing project. British ability to read German codes was very important in winning the war. Afterward, Britain faced the threat of Stalin's Soviet Union, and the Soviets did a great deal of spying. So at the time it must have seemed quite prudent to prevent the Soviets from learning about Britain's code-breaking experience and expertise.
Of course, later development of electronic computing technology would make that decision irrelevant, since encryption and decryption machines could be made far more compact and powerful than the German and British wartime technologies. But who in the late 1940s could have known that would happen? The transistor wasn't invented until 1947, and wasn't used widely in computers until the late 1950s. Integrated circuits were invented in 1958 and didn't enter computers until the 1960s. In 1946 no one had any good reason to think that the Colossus experience might not be useful later; thus it was worth protecting.
And Britain didn't really pay a price for keeping Colossus secret, apart from limited bragging rights, which don't mean that much anyway. Britons learned a lot from the American ENIAC experience and were just as active as the Americans in building early computers. The first few stored-program computers were all British, as was the first commercial computer (the Ferranti Mark 1, which was introduced just before the first UNIVAC). So British computer development didn't need Colossus to move ahead, and probably would not have advanced much by knowing about it.
Of course, the British position in the computer industry declined over time, but this has much more to do with the U.S. simply being far larger and far more economically powerful. (Perhaps Britain's postwar socialist policies had an effect as well, I don't know.)
You're right, a lot depends on how you define "computer". The early to mid 1940s saw several electromechanical and electronic digital computing machines in the U.S., Britain, and Germany: the Atanasoff-Berry computer, the Harvard Mark I, the ENIAC, the Colossus, and Konrad Zuse's Z3 and Z4.
I'm going to propose something a bit heretical. When I consider what "computer" means to us now, I think these early machines might be better called "proto-computers". The word "computer" makes me think of a fully electronic, digital machine that can do general-purpose computations, so that rules out electromechanical relay-based machines like the Zuse machines and the Harvard Mark I, as well as electronic machines specialized for one specific use like the Colossus and the Atanasoff-Berry computer. I also think of a machine that stores its programs in the same sort of electronic memory used for the data. This excludes the ENIAC, which was programmed by rewiring.
In other words, to me "computer" really is a synonym for what was originally called a "general-purpose electronic digital stored-program computer".
That type of computer was created after World War II. Around the time ENIAC was completed, its creators and other Americans and Britons interested in electronic computing discussed how to design a computer that would run stored programs. Several papers were published based on these discussions.
Afterward, many of these people rushed off to build their own machines. The British were the first to complete theirs, followed by Americans and Australians. In my view, these were the first true computers, according to the modern meaning of the term. Since their development was really a group effort, I don't think it means very much for any one nation or one machine to claim special credit for being first here.
By the way, Colossus also has a special disadvantage in taking credit for being an important step in the development of computers. Since it was a top secret project, the engineers who worked on it could only apply their experience indirectly, and others could not learn anything from its construction. So Colossus could not contribute as much to computer design as it otherwise might have.
Slashdot Mirror
They also used systems based on signal delays through wire and through quartz. (The wire system is also described on the Wikipedia page.)
The main memory held 1000 words, but that was composed of 100 mercury channels. (The whole memory system had 126 channels, but some of the channels were not part of main memory.) Each channel was a single acoustic path through a mercury tank, with 18 channels per tank. A main memory channel stored 10 words, with each word composed of 12 characters (11 digits plus sign), each of which was 6 bits plus a parity bit. That comes out to 840 bits.
The note estimates 1000 total bits per channel based on the clock rate and the maximum access time, but I think he got his numbers a little wrong. (He used a clock rate of 2.5 MHz, but it should be 2.25 MHz, which gives an answer of 900 bits.)
How about old enough to run for President?
One of the submitters to this article got Warren Toomey's name wrong.
You're half right there. The first implementation of Unix was in 1969 on a PDP-7, not a PDP-10. The two machines are quite different: the PDP-7 was a small 18-bit minicomputer designed for small groups of users, while the PDP-10 was a large 36-bit mainframe-type machine designed to serve big communities, like an entire university.
It was natural for Thompson and Ritchie to move from the PDP-7 to the 16-bit PDP-11, since the two machines were of similar scale in the beginning. Moving from the PDP-10 to the PDP-11 would have been like moving from your nice roomy house into your college friend's attic. :-)
There were a number of interesting operating systems on the PDP-10 (TENEX, TOPS-10, TOPS-20), but no one has ever implemented a Unix-style OS on it, as far as I know.
Dennis Ritchie's web site has a good account of the early development of Unix. It makes interesting reading if you're curious about such things.
Unfortunately, the joke of Ford Prefect's name was lost on most Americans, myself included. I didn't even know the car existed until I read the obituary for Douglas Adams, and my dad (an Englishman old enough to remember the Prefect) mentioned the car.
Wikipedia has a pretty good entry covering both the car and Adams' choice of the name for his character.
The biggest problems with SPEC's CPU benchmarks is that they tend to concentrate on technical applications and that people only talk about the average SPECint and SPECfp scores, neglecting the individual benchmark scores that correspond to real tasks. But you can always find the individual benchmark scores on SPEC's website.
Yeah, I suppose the body of the paper is really aimed at other geophysicists. Sorry about that. I didn't try to thoroughly comprehend all his calculations either, since I've left the field and haven't studied any of this stuff in about a decade. But I think I got the gist of his ideas. Let me try to answer your points.
1) The reason you use all that iron isn't to keep the blob hot. (Actually, it stays hot by itself; as it falls, the gravitational energy released from its fall heats it up. But this extra heat is conducted away to the surrounding rock, so the blob won't get too hot and melt the probe.) To reach the core, the iron is pushing on the bottom edge of a vertical crack, forcing it open all the way down. But to force the crack open, you need a lot of weight. Hence the big mass of iron. A blob of radioactive material would have to be almost as big to do the job, and would be a lot more expensive and...hard to work with. :-)
Reading that part of the paper again, I notice that he's referring to an existing, accepted theory used for understanding the movement of molten magma upward through cracks in the crust. It works in the same way, only upward (because the molten magma is lighter than the surrounding rock, unlike the iron here, which is heavier). The point is that this part of the idea is based on an understood phenomenon.
2) I think that comment was meant to be tongue-in-cheek, just like the paper title.
3) Well, the crack closes up, but there's still going to be a narrow fracture where the crack was. The crack isn't going to instantly heal, since mantle rock at that depth and pressure is solid. And there might also be a small amount of iron left behind in the crack. As he said in his annotation, "This may be a problem with the whole idea."
4) I think the idea here is mostly to investigate the outer core, not the intervening mantle. (As I recall, the mantle is much better known than the core.) So he doesn't care about measurements until the probe reaches the core. Once the liquid blob reaches the liquid outer core, it will presumeably spread out and mix somehow, while the probe, being solid, will keep sinking for a bit, long enough to function for a while with appropriate engineering. :-) This is where things get fuzzy for me; since the probe was built to be neutrally buoyant in the blob, how does it cease to float around in blob material once it gets to the core? Would its material have to be chosen to change density at the proper pressure so that it would start to sink?
5) Actually, he's really saying that we don't know much about how the intervening material distorts a seismic signal. But the neat thing about digital signals is that as long as you can read the bits, you don't care too much about the shape of the waveform. That's not true when "imaging" earth's interior using the travel times of earthquake seismic waves. Also, the idea is to send the data using compressional waves, which travel through liquid just fine. Of course, you're right that you'd have to get lots of practice transmitting data seismically first.
Naturally, I realize this idea is heavily on the loony side. That's what made it fun to read and think about. :-)
Magnetic drums weren't used for quite the same purposes as disks, though. Disks were for file storage, but drums were more often used as a low speed high-density working memory, similar to a modern virtual memory swap device. Running programs were swapped between core memory and the drum.
Before magnetic core memory became cheap enough to use in low-cost computers, some people even built low-end computer systems with a magnetic drum as the only working memory. Such computers were much slower than computers with electronic memory, since you might have to wait an entire rotational delay for a desired memory word to come under the read heads. Clever programmers arranged the instructions and data on the drum memory to minimize this delay.
Remember, Colossus was developed as a code-breaking enterprise, not as a computing project. British ability to read German codes was very important in winning the war. Afterward, Britain faced the threat of Stalin's Soviet Union, and the Soviets did a great deal of spying. So at the time it must have seemed quite prudent to prevent the Soviets from learning about Britain's code-breaking experience and expertise.
Of course, later development of electronic computing technology would make that decision irrelevant, since encryption and decryption machines could be made far more compact and powerful than the German and British wartime technologies. But who in the late 1940s could have known that would happen? The transistor wasn't invented until 1947, and wasn't used widely in computers until the late 1950s. Integrated circuits were invented in 1958 and didn't enter computers until the 1960s. In 1946 no one had any good reason to think that the Colossus experience might not be useful later; thus it was worth protecting.
And Britain didn't really pay a price for keeping Colossus secret, apart from limited bragging rights, which don't mean that much anyway. Britons learned a lot from the American ENIAC experience and were just as active as the Americans in building early computers. The first few stored-program computers were all British, as was the first commercial computer (the Ferranti Mark 1, which was introduced just before the first UNIVAC). So British computer development didn't need Colossus to move ahead, and probably would not have advanced much by knowing about it.
Of course, the British position in the computer industry declined over time, but this has much more to do with the U.S. simply being far larger and far more economically powerful. (Perhaps Britain's postwar socialist policies had an effect as well, I don't know.)
I'm going to propose something a bit heretical. When I consider what "computer" means to us now, I think these early machines might be better called "proto-computers". The word "computer" makes me think of a fully electronic, digital machine that can do general-purpose computations, so that rules out electromechanical relay-based machines like the Zuse machines and the Harvard Mark I, as well as electronic machines specialized for one specific use like the Colossus and the Atanasoff-Berry computer. I also think of a machine that stores its programs in the same sort of electronic memory used for the data. This excludes the ENIAC, which was programmed by rewiring.
In other words, to me "computer" really is a synonym for what was originally called a "general-purpose electronic digital stored-program computer".
That type of computer was created after World War II. Around the time ENIAC was completed, its creators and other Americans and Britons interested in electronic computing discussed how to design a computer that would run stored programs. Several papers were published based on these discussions.
Afterward, many of these people rushed off to build their own machines. The British were the first to complete theirs, followed by Americans and Australians. In my view, these were the first true computers, according to the modern meaning of the term. Since their development was really a group effort, I don't think it means very much for any one nation or one machine to claim special credit for being first here.
By the way, Colossus also has a special disadvantage in taking credit for being an important step in the development of computers. Since it was a top secret project, the engineers who worked on it could only apply their experience indirectly, and others could not learn anything from its construction. So Colossus could not contribute as much to computer design as it otherwise might have.