Your compiler has not to be entirely bugfree. It has just to be good enough to compile another compiler correctly whose source code seems to be backdoor free. So then you have a compiler which should be good enough to compile a tool chain for you.
What I wanted to point out in my former post is to guarrant that your computer gets infected you have to put a backdoor in everything running on your computer. As long as every piece of code can be replaced at will, the infection chain is due to break. It is much too fragile to survive for a longer time without knowledge of the maintainers. FOSS has this great feature: Even though parts of the system may get compromised, this may not survive for long under the radar screen of the maintainers.
A similar problem occurs in the so called copy protection schemes. They work as long as every piece of code touching the data in question is cooperative. As soon as a single part breaks the rules the whole copy protection is void. This would be one of the main reasons to make breaking those schemes illegal, because the technical warranties are virtually nonexistant for any copy protection scheme to survive.
So I bought my car five years ago with 81kW (113 HP), and I could have bought a similar car for the same price with an 162kW engine 42 month ago, and half a year ago I should have been able to get a 32-piston-648kW (904 HP) car for the same money?
But then your compiler has to determine that the code you are compiling is actually a netfilter (or similar) code. Otherwise the compiler has to put in the backdoor in about every piece of software it is compiling. And then the backdoor should reside in a part of code that is quite probable to be executed during the program's lifetime.
Mr. Kernighan's "backdoor compiler" is all well and neat as a theoretical concept, but the reality is something different:) To realize it you have to put a lot of artificial intelligence into your compiler, and you should make sure no one ever wants to profile the code it is compiling, otherwise it will show up in the debugger. And infecting the whole tool chain from compilers and script languages up to profiles, debuggers and trace programs could prove quite complicated even for a determined party.
I used the 'ies' form, because it's actually not "eightys" but "eighties";). So the correct form should be '80(-y)ies', but this is a letter longer and uses two more shifts than the word 'eighties' written fully in letters, so the saving in typing is lost. But who am I to argue anyway, me not being a native speaker?
Don't underestimate the storage needed for application development. With all this versioning, where you keep the former version of every change and every nightly build, with testing scenarios (and their own versioning) and testing protocols the space needed can easily be a thousand times the space of the actual application.
...but the point is Office just can't handle anything that wasn't originally created by MS.
So, is that because of incompetence, or by design?
It's by design. When MS Word was being pushed by Microsoft as "industry standard" (back in the late '80ies, early '90ies), it came with dozens of import filters for about any word processor format known to Man. So the MS sales person could always point out that no one would loose any old data, because Word was pretty capable of reading the format in question.
With the later versions, the number of file formats MS Word was supporting, shrank. And today it is reduced to old MS Word formats (and none of them as perfect as other office suites) and to a number of good documented formats (RTF, HTML, plain text). I remember when the company I was working for was converting from OS/2 to Windows NT4.0 and the old Ami Pro documents were no longer readable. It was quite an effort to finally find an old copy of Winword 6.0a to import the Ami Pro files, because the later incarnations of MS Word weren't able to read them directly.
With BitTorrent you are forced to share. So when you start downloading something you are also sharing copyrighted material.
You fell for a small but important trap. Just because you are downloading something doesn't mean you are downloading copyrighted material. And just because you are downloading copyrighted material it doesn't mean you are doing anything illegal. And just because you are downloading copyrighted material and at the same moment share it to other people interested in the same material you are still not doing anything illegal.
Many people distribute their own stuff with BitTorrent, because so they only need to seed the material and have maybe to handle the amount for two or three complete downloads from their site. All the other people get their data from those who already got the block in question. This is a huge bandwith saver for all involved parties. Mandrake Linux gets distributed that way for instance, and you can also get other distributions that way, completely legal.
So different than KaZaA or other file sharing networks where you have to look out to find some legitimate use, with BitTorrent it's easy.
On the other hand: The article especially points out that one of the counter technics employed by MPAA and RIAA is to put bogus files with an interesting file name up for sharing. So people searching for a special file may end up with a file with the same name but a different content. Centralized hosting of "proven" files like Napster did made them liable for copyright infringement. Calculating checksums would only help if you could compare the checksums with a trusted database, which is open to the same type of liability, because the database has to calculate the checksums by using the original files. So infecting KaZaA with bogus files and forged checksums is easy. And you know how many files everyone is actually sharing, because you can just query the client and ask. This makes KaZaA and similar systems vulnerable to two types of attacks: Tracking people distributing immense amounts of files, and poisoning the data pool with fake files.
With BitTorrent it's different. Everyone seeding a file is taking technical responsibility for the correctness of the file. But the actual data blocks are coming from other computers. The tracker keeps track of the different computers sharing exactly this file. For every other file there is another tracker. So with the tracker data you can actually find out who's sharing a specified single file, and because of checksumming you can be confident that all people listed in the tracker are sharing the same file derived from the same original data. But you don't know which files else are being shared on the same computer, because their trackers are being somewhere else. So all a copyright infringement tracking bot connecting to that tracker gets to know is information about this single file. Quite ineffective, and if you go to a judge with an IP address and tell them: "From this IP this one file was distributed" he'll probably tell you that he has more important stuff to do.
So people sharing large amounts of data don't appear any different to the bot than people just sharing a single file. And poisoning the data is also not easy, because you can't spoof someone elses seed and tracker. So you have to establish yourself as a second source, with your own seed and tracker, and if you don't provide correct data, your seed will die out soon. There may be an attack possible if you hack BitTorrent yourself (it's GPLed, after all), and if you take part of the actually ongoing BitTorrent sharing, but send fake blocks to every request you get. But then the checksumming will detect you quite soon, and you loose again.
What's really wrong with taking the Constitution literally and have the jury really be a party of peers, that means people with about the same background than the defendant? When the "jury of peers" was introduced, it meant that an aristocrat could only be judged by a jury consisting of aristocrats, and a free man only by a jury of free men. In a society where everyone should be equal before the law, the jury of peers has lost its deeper meaning: A jury of people from your milieu, class or group. So why not having a jury of engineers for patent infringment cases?
Simple point is that a scientific consensus has not always been the most accurate thing in the world.
Simple point is, that science in the way we know it today is a thing developed during the 19th century. Before it there was knowledge, there was lots of speculation and there was no separation between hypothesis, theory and fact.
Even the great scientist of the late 17th and early 18th century, Isaac Newton, was more interested in alchemic experiments and metaphysical speculation than in the Physics and Mathematics where he laid the fundamentations for today's Calculus and Mechanics.
Today's scientific consensus is still lacking, and it is no replacement for a thoroughly tested theory. But it is the best thing we can get, if said theory is still missing. Just because it's not perfect you shouldn't just throw it away.
First: Germany fares quite low in the table. Second: The testing sample in Germany consisted of pupils of all types of schools. Third: It is currently discussed in Germany if separating the children at age 10 is way to early, because then the further way in life is somehow determined, and the motivation to learn is taken from everyone. People attending the Gymnasium think they are fine off anyway, because they get higher education without further effort, people at the Realschule think, fine, so they will learn a trade, no point in looking for higher math skills or developing an interest in literature. And people at the Hauptschule think, they are losers anyway, so what's the point in learning?
No, in German, there are even two different conjunctives. The one is for "if" cases, the other one for the so called "indirect speech", when you quote something someone else said. "Er sagt, er haette es schon erledigt." (he says he did it already) is different than "Er sagt, er hatte es schon erledigt." (note the Umlaut 'ae' in 'haette' vs. the 'a' in 'hatte'.) The second one is only used in casual speech, where people often don't use the conjunctive.
Old Greek (I don't know about the modern Greek) has a similar differentiation. There is one conjunctive that's used if you are talking about possibilities, and the other one you use if you point out something that's impossible.
How many users actually use the VB support in Office? I'll bet it's under one percent.
In every company I worked so far, especially Microsoft Excel was used heavily relying on the builtin VB. In fact it's one of the real strengths of Excel.
Making a computer safe is hard. But for most people, a computer is just a word processor/web browser/email client. Making that safe is certainly doable. Why hasn't anyone done it yet? That's my complaint.
Because there is currently no way to make a restricted computer actually cheaper than the fully fledged one. There have been several attempts at reducing the capabilities of computers for simple home/office use. Remember the NetPC, the Thin Client, the JavaStations? None of them worked. They weren't cheaper than the all-purpose computer (because they basicly rely on the same hardware), and they were less flexible.
It is in fact much more expensive to build a sufficiently restricted computer than an all-purpose one. An all-purpose computer just needs a Turing complete CPU (a small microprocessor with the OpCodes INC, DEC and JZ is Turing complete), sufficient memory and I/O interfaces. A restricted computer needs much more than that. Either it has a Turing complete core and then lots of assistance checks&bounds builtin, or it uses a not really Turing complete core, where all OpCodes are limited in a sophisticated way.
And now comes the biggest disadvantage: An all-purpose computer can emulate the restricted one (given the right software), but not vice versa. So in the end the all-purpose computer prevailed: It could do everything the restricted computer was capable of at a cheaper price. And the real maintenance cost (TCO) is still much debated and complete meaningless to people, who own a computer in private and just call the boy from the neighborhood, who is so intelligent and good with that thing.
Here's how to stop a computer needing maintenance, ever: put all the programs you're going to need in ROM. Clear RAM whenever you reboot.
There. That wasn't hard, was it?
Yes. No persistant data storage. No way to actually create new programs. No way to use remote ressources. No protection for so called active content (program builtin languages) not running havoc. What you create is a quite limited type of computer, similar to a game console or an early '80ies home computer without external storage.
We are talking about computers you actually want to work with.
So now lets talk about the real components such a computer needs: verified hardware (where the correct implementation is mathematically proved), verified compilers, verified operating system, verified applications, verified protocols for remote usage.
Four out of those five requirements are already accomplished or on its way to accomplishment. There is hardware (CPU, memory...) where there is a mathematical proof that the implementation is an actual representation of the specification. There are verified compilers where there is mathematical proof, that the object code they put out is mathematically equivalent to the source code you are feeding it. There is work in progress to prove the correctness of an operating system (I should check with my old operating system group if they are finished yet). There are lots of network protocols whose correctness is proved for both the protocol itself and an actual implementation of the protocol.
So there is one big block remaining: verified applications. And there we are back at Step 1. No one hinders us to implement a hardware layer or a complete operating system at application level (You don't believe me? Look at VirtualPC [OS] or VMware [hardware layer]).
Any application that has a Turing complete subsystem (like most Office suits with their application specific languages) can be host system for the same thing: You could even create a Linux being hosted by Microsoft Word (write an C Compiler in VBA and then port the Linux kernel. Simulate a simple framebuffer device on the Word canvas in VBA and port X11 etc.pp.).
So having verified applications still doesn't warrant a maintenance free computer. Even the data the applications get feed with has to be verified. And that's the point where a computer turns from a useful tool into a completely verified and maintenance free but even so completely unusable piece of junk, because you have to mathematically prove the correctness of your own data.
There is no technical reason why a computer should need any maintenance at all.
There is one. It consists of two components. The first one is called Turing completeness, the second one is called Termination problem.
If you have an object that is Turing complete, then there is no way to prove for sure that it will execute exactly according to the specification, because you can't even prove (in general) that it will ever stop with whatever it is doing. This is founded deep down in the Mathematics of computing, so there is no way to engineer around.
So if you can't be sure, you have to do maintenance to make sure it will perform as expected, because then the computer needs manual supervision to externally stop programs from running for eternity or doing arbitrary things you never expected them to do. Computers are per definition all-purpose tools, and they can be modified to do things that cover many different areas of knowledge and science. No single person and not even a larger team is able to oversee what a computer is capable of, because of this general openess. You can only limit the damage a computer is able to do by limiting the ways it is able to communicate with its exterior, because for now computers are limited to logical operations, and they need external hardware to have a physical impact to their environments. On the other hands: What we need from computers are mostly not logical operations, but interactions with external hardware to have this physical impact to the surrounding world, may it be printing out a document or sending an electrical signal to another computer, changing the color of pixels on a screen or reading the buffer of a keyboard connected to it.
So if no one can specify the limits of what a computer may be able to do, no one can say for sure how to limit the damages a computer can do. There is only one way to minimize the impact: continously monitoring the computer while it is running and correct everything that goes wrong. This is called maintenance.
You can limit a computer and its hardware to a certain subset of possible operations, but then you end up with an electronic dish washer, a phone headset, a CD player or any other electronic gadget. But not with an all-purpose computer.
It is useful if they find other things you print (maybe with your sender address imprinted on it), and can compare them. Often they have already a suspicion, and all they need is a confirmation. So why not trying to get hold of something innocent you just printed which can be easily tracked back to you?
Of course the confound is that every country has a special set of rules to do with money and it becomes a special case. (e.g. rules about how damaged/what kind of damage a bill can have before it is void/worth half as much etc) But that is the general idea as far as I understand it.
That's the main difference between a bill and a battery. The value of a battery degrades slowly, until it becomes completely unusable. The value of a bill stays the same until it gets rejected by the central bank, then the value suddenly drops to zero. You could speculate with the probability of the bill being rejected and create a kind of synthetic value degradation, where the value of the bill is represented by the expected rejection rate of similarly used and destroyed bills, but this is a rather theoretic way to smoothen the "value jump". For our model the jump from value = 1 to value = 0 fits better.
The threat level is reached in the server environment. The growth of the Windows Server offerings was planned to take up all the market share the UNIX companies were losing. But they don't loose to Windows, they loose to Linux. So basicly Microsoft can't run into the UNIX world as they planned, but Linux is walking in. 10 years ago the calculation was that Windows NT offers the migration path away from UNIX to a Microsoft world (Windows NT is certified UNIX95 compliant, even though in 1995, Microsoft owned the UNIX trademark;) ). As it seems Windows NT and the subsequent versions have to create their own market (they often grow out of already installed Windows ecologies with a need to centralize services), and Linux is taking over the market for companies searching for a cheap alternative to UNIX offerings.
The problem is: The U.S. seem to make decisions based on exactly those parodies of itself. So other countries can't even dare to laugh, they have to brace for impact.
Ok... slowly again: The method is about getting a huge, heavy object with welldefined edges into a swing. You can do that with a washing machine on your own (because it barely weighs about 500 pounds) or you take 20 or 50 people and try it at a stone block of about 10 or 20 tons. Thor Heyerdahls team had ropes bound on the top of the stone head and was pulling it right and left until it swung slowly. A method to erect such a massive statue was tried before already (lifting it with wooden levers inch for inch and putting small stones and earth in the opening gap). Then they were directing the swing to get the stone moving slowly.
Btw.: The first Heyerdahl expedition (1955/1956) lead to the edition of the book "Aku-Aku. The secrets of Easter Island". The moving experiment was performed at his second large Easter Island expedition 1986-1988.
And we should never forget: Those living 10,000 years ago were already Homo sapiens sapiens, so no less capable in thinking than we are today. Just because a group of archeologists can't think of a sufficient technology on the spot to fulfill a task humans fulfilled thousands of years ago doesn't mean that such a technology couldn't have existed then. Archeologists need a broad knowledge of different technologies, but so they often have to sacrify deepness. Often the appearent mystery solves if you have some non-archeologist look at the site, who is a specialist on a certain technology.
Thor Heyerdahl (ok, no archeologist at all, but an ethnograph) once tried to move one of the statues from the Easter Island out of the quarry it was still in, and all the methods he was trying failed, until one of the local people told him that the statues 'walked upright'. Finally an engineer showed him how you get a statue walk upright: Have it swing slightly from right to left, so one of the edges is on the ground and the other one is lifted. Then you can turn it slightly so that in the next swing the lifted edge lands a little bit further down the road. From a distance it looks as if the statue is swaggering forward. (I was using a similar method to get my washing machine into my house:), works like a charm, and you can do it for yourself, without too much help from others.) And the local people were glad finally someone understood what they were trying to tell them from the beginning. If Thor Heyerdahl had asked any moving contractor before, he would have been done long ago;).
Often it's not the technology level itself that shows how advanced a civilisation is. Often it's all those little tricks, which use existing technology to its maximum, that makes a civilisation more successful than others. The middle age guilds of craftmanship were basicly competence centers which gave all those little tricks to the next generation of craftsmen. Just by looking at the tools you can't really say what was possible for a craftsman then.
And for the age of Stonehenge: The latest additions to Stonehenge were done around 1500 B.C., when all egyptian pyramids were already built long ago (pyramids as king's tombs were abandoned around 1900 B.C. with the demise of the XIth dynasty), and when the SkyDiscofNebra (~1600 B.C.) was already cast in bronze.
Slashdot Mirror
A God surely can. :) But try to tell a creationist, that his God has evolved a little since the Creation :)
Your compiler has not to be entirely bugfree. It has just to be good enough to compile another compiler correctly whose source code seems to be backdoor free. So then you have a compiler which should be good enough to compile a tool chain for you.
What I wanted to point out in my former post is to guarrant that your computer gets infected you have to put a backdoor in everything running on your computer. As long as every piece of code can be replaced at will, the infection chain is due to break. It is much too fragile to survive for a longer time without knowledge of the maintainers. FOSS has this great feature: Even though parts of the system may get compromised, this may not survive for long under the radar screen of the maintainers.
A similar problem occurs in the so called copy protection schemes. They work as long as every piece of code touching the data in question is cooperative. As soon as a single part breaks the rules the whole copy protection is void. This would be one of the main reasons to make breaking those schemes illegal, because the technical warranties are virtually nonexistant for any copy protection scheme to survive.
So I bought my car five years ago with 81kW (113 HP), and I could have bought a similar car for the same price with an 162kW engine 42 month ago, and half a year ago I should have been able to get a 32-piston-648kW (904 HP) car for the same money?
But then your compiler has to determine that the code you are compiling is actually a netfilter (or similar) code. Otherwise the compiler has to put in the backdoor in about every piece of software it is compiling. And then the backdoor should reside in a part of code that is quite probable to be executed during the program's lifetime.
:) To realize it you have to put a lot of artificial intelligence into your compiler, and you should make sure no one ever wants to profile the code it is compiling, otherwise it will show up in the debugger. And infecting the whole tool chain from compilers and script languages up to profiles, debuggers and trace programs could prove quite complicated even for a determined party.
Mr. Kernighan's "backdoor compiler" is all well and neat as a theoretical concept, but the reality is something different
I used the 'ies' form, because it's actually not "eightys" but "eighties" ;). So the correct form should be '80(-y)ies', but this is a letter longer and uses two more shifts than the word 'eighties' written fully in letters, so the saving in typing is lost. But who am I to argue anyway, me not being a native speaker?
Don't underestimate the storage needed for application development. With all this versioning, where you keep the former version of every change and every nightly build, with testing scenarios (and their own versioning) and testing protocols the space needed can easily be a thousand times the space of the actual application.
It's by design. When MS Word was being pushed by Microsoft as "industry standard" (back in the late '80ies, early '90ies), it came with dozens of import filters for about any word processor format known to Man. So the MS sales person could always point out that no one would loose any old data, because Word was pretty capable of reading the format in question.
With the later versions, the number of file formats MS Word was supporting, shrank. And today it is reduced to old MS Word formats (and none of them as perfect as other office suites) and to a number of good documented formats (RTF, HTML, plain text). I remember when the company I was working for was converting from OS/2 to Windows NT4.0 and the old Ami Pro documents were no longer readable. It was quite an effort to finally find an old copy of Winword 6.0a to import the Ami Pro files, because the later incarnations of MS Word weren't able to read them directly.
With BitTorrent you are forced to share. So when you start downloading something you are also sharing copyrighted material.
You fell for a small but important trap. Just because you are downloading something doesn't mean you are downloading copyrighted material. And just because you are downloading copyrighted material it doesn't mean you are doing anything illegal. And just because you are downloading copyrighted material and at the same moment share it to other people interested in the same material you are still not doing anything illegal.
Many people distribute their own stuff with BitTorrent, because so they only need to seed the material and have maybe to handle the amount for two or three complete downloads from their site. All the other people get their data from those who already got the block in question. This is a huge bandwith saver for all involved parties. Mandrake Linux gets distributed that way for instance, and you can also get other distributions that way, completely legal.
So different than KaZaA or other file sharing networks where you have to look out to find some legitimate use, with BitTorrent it's easy.
On the other hand: The article especially points out that one of the counter technics employed by MPAA and RIAA is to put bogus files with an interesting file name up for sharing. So people searching for a special file may end up with a file with the same name but a different content. Centralized hosting of "proven" files like Napster did made them liable for copyright infringement. Calculating checksums would only help if you could compare the checksums with a trusted database, which is open to the same type of liability, because the database has to calculate the checksums by using the original files. So infecting KaZaA with bogus files and forged checksums is easy. And you know how many files everyone is actually sharing, because you can just query the client and ask. This makes KaZaA and similar systems vulnerable to two types of attacks: Tracking people distributing immense amounts of files, and poisoning the data pool with fake files.
With BitTorrent it's different. Everyone seeding a file is taking technical responsibility for the correctness of the file. But the actual data blocks are coming from other computers. The tracker keeps track of the different computers sharing exactly this file. For every other file there is another tracker. So with the tracker data you can actually find out who's sharing a specified single file, and because of checksumming you can be confident that all people listed in the tracker are sharing the same file derived from the same original data. But you don't know which files else are being shared on the same computer, because their trackers are being somewhere else. So all a copyright infringement tracking bot connecting to that tracker gets to know is information about this single file. Quite ineffective, and if you go to a judge with an IP address and tell them: "From this IP this one file was distributed" he'll probably tell you that he has more important stuff to do.
So people sharing large amounts of data don't appear any different to the bot than people just sharing a single file. And poisoning the data is also not easy, because you can't spoof someone elses seed and tracker. So you have to establish yourself as a second source, with your own seed and tracker, and if you don't provide correct data, your seed will die out soon. There may be an attack possible if you hack BitTorrent yourself (it's GPLed, after all), and if you take part of the actually ongoing BitTorrent sharing, but send fake blocks to every request you get. But then the checksumming will detect you quite soon, and you loose again.
What's really wrong with taking the Constitution literally and have the jury really be a party of peers, that means people with about the same background than the defendant? When the "jury of peers" was introduced, it meant that an aristocrat could only be judged by a jury consisting of aristocrats, and a free man only by a jury of free men. In a society where everyone should be equal before the law, the jury of peers has lost its deeper meaning: A jury of people from your milieu, class or group. So why not having a jury of engineers for patent infringment cases?
Simple point is that a scientific consensus has not always been the most accurate thing in the world.
Simple point is, that science in the way we know it today is a thing developed during the 19th century. Before it there was knowledge, there was lots of speculation and there was no separation between hypothesis, theory and fact.
Even the great scientist of the late 17th and early 18th century, Isaac Newton, was more interested in alchemic experiments and metaphysical speculation than in the Physics and Mathematics where he laid the fundamentations for today's Calculus and Mechanics.
Today's scientific consensus is still lacking, and it is no replacement for a thoroughly tested theory. But it is the best thing we can get, if said theory is still missing. Just because it's not perfect you shouldn't just throw it away.
First: Germany fares quite low in the table.
Second: The testing sample in Germany consisted of pupils of all types of schools.
Third: It is currently discussed in Germany if separating the children at age 10 is way to early, because then the further way in life is somehow determined, and the motivation to learn is taken from everyone. People attending the Gymnasium think they are fine off anyway, because they get higher education without further effort, people at the Realschule think, fine, so they will learn a trade, no point in looking for higher math skills or developing an interest in literature. And people at the Hauptschule think, they are losers anyway, so what's the point in learning?
No, in German, there are even two different conjunctives. The one is for "if" cases, the other one for the so called "indirect speech", when you quote something someone else said. "Er sagt, er haette es schon erledigt." (he says he did it already) is different than "Er sagt, er hatte es schon erledigt." (note the Umlaut 'ae' in 'haette' vs. the 'a' in 'hatte'.) The second one is only used in casual speech, where people often don't use the conjunctive.
Old Greek (I don't know about the modern Greek) has a similar differentiation. There is one conjunctive that's used if you are talking about possibilities, and the other one you use if you point out something that's impossible.
You won't throw out antiques, won't you?!
Collectibles, I say! Collectibles!
Because sometimes it requires some insight to make such a funny comment?
How many users actually use the VB support in Office? I'll bet it's under one percent.
In every company I worked so far, especially Microsoft Excel was used heavily relying on the builtin VB. In fact it's one of the real strengths of Excel.
Making a computer safe is hard. But for most people, a computer is just a word processor/web browser/email client. Making that safe is certainly doable. Why hasn't anyone done it yet? That's my complaint.
Because there is currently no way to make a restricted computer actually cheaper than the fully fledged one. There have been several attempts at reducing the capabilities of computers for simple home/office use. Remember the NetPC, the Thin Client, the JavaStations? None of them worked. They weren't cheaper than the all-purpose computer (because they basicly rely on the same hardware), and they were less flexible.
It is in fact much more expensive to build a sufficiently restricted computer than an all-purpose one. An all-purpose computer just needs a Turing complete CPU (a small microprocessor with the OpCodes INC, DEC and JZ is Turing complete), sufficient memory and I/O interfaces. A restricted computer needs much more than that. Either it has a Turing complete core and then lots of assistance checks&bounds builtin, or it uses a not really Turing complete core, where all OpCodes are limited in a sophisticated way.
And now comes the biggest disadvantage: An all-purpose computer can emulate the restricted one (given the right software), but not vice versa. So in the end the all-purpose computer prevailed: It could do everything the restricted computer was capable of at a cheaper price. And the real maintenance cost (TCO) is still much debated and complete meaningless to people, who own a computer in private and just call the boy from the neighborhood, who is so intelligent and good with that thing.
Here's how to stop a computer needing maintenance, ever: put all the programs you're going to need in ROM. Clear RAM whenever you reboot.
There. That wasn't hard, was it?
Yes. No persistant data storage. No way to actually create new programs. No way to use remote ressources. No protection for so called active content (program builtin languages) not running havoc. What you create is a quite limited type of computer, similar to a game console or an early '80ies home computer without external storage.
We are talking about computers you actually want to work with.
So now lets talk about the real components such a computer needs: verified hardware (where the correct implementation is mathematically proved), verified compilers, verified operating system, verified applications, verified protocols for remote usage.
Four out of those five requirements are already accomplished or on its way to accomplishment. There is hardware (CPU, memory...) where there is a mathematical proof that the implementation is an actual representation of the specification. There are verified compilers where there is mathematical proof, that the object code they put out is mathematically equivalent to the source code you are feeding it. There is work in progress to prove the correctness of an operating system (I should check with my old operating system group if they are finished yet). There are lots of network protocols whose correctness is proved for both the protocol itself and an actual implementation of the protocol.
So there is one big block remaining: verified applications. And there we are back at Step 1. No one hinders us to implement a hardware layer or a complete operating system at application level (You don't believe me? Look at VirtualPC [OS] or VMware [hardware layer]).
Any application that has a Turing complete subsystem (like most Office suits with their application specific languages) can be host system for the same thing: You could even create a Linux being hosted by Microsoft Word (write an C Compiler in VBA and then port the Linux kernel. Simulate a simple framebuffer device on the Word canvas in VBA and port X11 etc.pp.).
So having verified applications still doesn't warrant a maintenance free computer. Even the data the applications get feed with has to be verified. And that's the point where a computer turns from a useful tool into a completely verified and maintenance free but even so completely unusable piece of junk, because you have to mathematically prove the correctness of your own data.
There is no technical reason why a computer should need any maintenance at all.
There is one. It consists of two components. The first one is called Turing completeness, the second one is called Termination problem.
If you have an object that is Turing complete, then there is no way to prove for sure that it will execute exactly according to the specification, because you can't even prove (in general) that it will ever stop with whatever it is doing.
This is founded deep down in the Mathematics of computing, so there is no way to engineer around.
So if you can't be sure, you have to do maintenance to make sure it will perform as expected, because then the computer needs manual supervision to externally stop programs from running for eternity or doing arbitrary things you never expected them to do. Computers are per definition all-purpose tools, and they can be modified to do things that cover many different areas of knowledge and science. No single person and not even a larger team is able to oversee what a computer is capable of, because of this general openess. You can only limit the damage a computer is able to do by limiting the ways it is able to communicate with its exterior, because for now computers are limited to logical operations, and they need external hardware to have a physical impact to their environments.
On the other hands: What we need from computers are mostly not logical operations, but interactions with external hardware to have this physical impact to the surrounding world, may it be printing out a document or sending an electrical signal to another computer, changing the color of pixels on a screen or reading the buffer of a keyboard connected to it.
So if no one can specify the limits of what a computer may be able to do, no one can say for sure how to limit the damages a computer can do. There is only one way to minimize the impact: continously monitoring the computer while it is running and correct everything that goes wrong. This is called maintenance.
You can limit a computer and its hardware to a certain subset of possible operations, but then you end up with an electronic dish washer, a phone headset, a CD player or any other electronic gadget. But not with an all-purpose computer.
It is useful if they find other things you print (maybe with your sender address imprinted on it), and can compare them. Often they have already a suspicion, and all they need is a confirmation. So why not trying to get hold of something innocent you just printed which can be easily tracked back to you?
Of course the confound is that every country has a special set of rules to do with money and it becomes a special case. (e.g. rules about how damaged/what kind of damage a bill can have before it is void/worth half as much etc) But that is the general idea as far as I understand it.
That's the main difference between a bill and a battery.
The value of a battery degrades slowly, until it becomes completely unusable. The value of a bill stays the same until it gets rejected by the central bank, then the value suddenly drops to zero. You could speculate with the probability of the bill being rejected and create a kind of synthetic value degradation, where the value of the bill is represented by the expected rejection rate of similarly used and destroyed bills, but this is a rather theoretic way to smoothen the "value jump". For our model the jump from value = 1 to value = 0 fits better.
If chimps do it with big sticks, it's natural. If humans do it with big sticks, it's called sodomy and considered against nature.
The threat level is reached in the server environment. The growth of the Windows Server offerings was planned to take up all the market share the UNIX companies were losing. But they don't loose to Windows, they loose to Linux. So basicly Microsoft can't run into the UNIX world as they planned, but Linux is walking in. ;) ). As it seems Windows NT and the subsequent versions have to create their own market (they often grow out of already installed Windows ecologies with a need to centralize services), and Linux is taking over the market for companies searching for a cheap alternative to UNIX offerings.
10 years ago the calculation was that Windows NT offers the migration path away from UNIX to a Microsoft world (Windows NT is certified UNIX95 compliant, even though in 1995, Microsoft owned the UNIX trademark
The problem is: The U.S. seem to make decisions based on exactly those parodies of itself. So other countries can't even dare to laugh, they have to brace for impact.
Ok... slowly again: The method is about getting a huge, heavy object with welldefined edges into a swing. You can do that with a washing machine on your own (because it barely weighs about 500 pounds) or you take 20 or 50 people and try it at a stone block of about 10 or 20 tons. Thor Heyerdahls team had ropes bound on the top of the stone head and was pulling it right and left until it swung slowly. A method to erect such a massive statue was tried before already (lifting it with wooden levers inch for inch and putting small stones and earth in the opening gap). Then they were directing the swing to get the stone moving slowly.
Btw.: The first Heyerdahl expedition (1955/1956) lead to the edition of the book "Aku-Aku. The secrets of Easter Island". The moving experiment was performed at his second large Easter Island expedition 1986-1988.
And we should never forget: Those living 10,000 years ago were already Homo sapiens sapiens, so no less capable in thinking than we are today. Just because a group of archeologists can't think of a sufficient technology on the spot to fulfill a task humans fulfilled thousands of years ago doesn't mean that such a technology couldn't have existed then. Archeologists need a broad knowledge of different technologies, but so they often have to sacrify deepness. Often the appearent mystery solves if you have some non-archeologist look at the site, who is a specialist on a certain technology.
:), works like a charm, and you can do it for yourself, without too much help from others.) And the local people were glad finally someone understood what they were trying to tell them from the beginning. If Thor Heyerdahl had asked any moving contractor before, he would have been done long ago ;).
Thor Heyerdahl (ok, no archeologist at all, but an ethnograph) once tried to move one of the statues from the Easter Island out of the quarry it was still in, and all the methods he was trying failed, until one of the local people told him that the statues 'walked upright'. Finally an engineer showed him how you get a statue walk upright: Have it swing slightly from right to left, so one of the edges is on the ground and the other one is lifted. Then you can turn it slightly so that in the next swing the lifted edge lands a little bit further down the road. From a distance it looks as if the statue is swaggering forward. (I was using a similar method to get my washing machine into my house
Often it's not the technology level itself that shows how advanced a civilisation is. Often it's all those little tricks, which use existing technology to its maximum, that makes a civilisation more successful than others. The middle age guilds of craftmanship were basicly competence centers which gave all those little tricks to the next generation of craftsmen. Just by looking at the tools you can't really say what was possible for a craftsman then.
And for the age of Stonehenge: The latest additions to Stonehenge were done around 1500 B.C., when all egyptian pyramids were already built long ago (pyramids as king's tombs were abandoned around 1900 B.C. with the demise of the XIth dynasty), and when the Sky Disc of Nebra (~1600 B.C.) was already cast in bronze.
You should'nt have clicked on the link, just copied the URL into your browser. Sheesh. Even the simplest gags still work around her.