As you know, the NT kernel is in a way a descendent of Digital's VMS. VMS in turned was based on a 16-bit operating system, RSX-11M. The interesting thing is that in the old days, propietary operating systems weren't particularly closed. RSX-11M was built from source code, so you always got the exec in source. Until VMS 4.5, the source listings were delivered on microfiche with the system. There were some very good books available on the internals of VMS too.
Personally, I think that the NT kernel seems quite ok. The trouble is that it is closed (so I can't easily understand what is going on) and it is surrounded by layers of crap. I agree that the Linux people could have learned a lot from the synchronisation objects in NT, but they could also have leaned a lot from older systems like VMS, which were extremely well documented. However, better asynch support is coming soon.
However, the next big area for Linux is resource locking (needed for clusters). NT does it, but not very well. VMS did it better (and still does, from last time I used it).
Iain Banks also has Orbitals (Ringworlds) in his Culture universe. The thing is aren't we talking about vehicles, which generally excludes B5, ringworlds and orbitals. A GSV is probably the largests actual moving space ship in current science fiction.
The point is that with TCPA, the hardware checks the signature on the software and the software allows only authorised updates. Great idea, and the principle has been discussed for a long time. The problem is that if there is any duff software in the trusted path then the system can be subverted. Then there is the issue of how to organise keys in a non-hierarchical environment.
They got rid of the VAX and put Alphas as hosts, but still running VMS and the host system behind Eurex is more COBOL than anything else. There is even a little FORTRAN there as well (a file hacking tool). Eurex is the world's biggest electronic derivatives exchange.
The problem with any system is that it can be compromised. If I do online banking and enter my access password, it may be captured in a number of different ways before it goes out to the bank. The problem is that the paassword can be discovered in a number of ways before the post goes via https to the bank's server.
Even when my password hits the https client software, how do I know that the information is really being sent securely? I don't.
The counter example used by the digital rights people is that when they send me a key to access controlled media, how can they be certain that I don't intercept the decoded bit stream?
In the first case, it is reasonable to have a trusted platform because the user can choose to accept what software he runs. In particular it can allow me to differentiate between an allowable update and one that isn't.
In the second, then then the owner/user of the system can not be permitted to have control. If the user is permitted to have full control then the platform must disclose to the access granter that the link between the media decryption engine and the output can no longer be trusted.
One can argue that the first is reasonable but the second would prevent anyone from looking at digitally licensed media on an open computing platform such as Linux.
In any case, this all supposes that the platform as installed, is indeed secure. It probably isn't. Even systems that implement a good security reference monitor can be compromised by poor configuration and software layers that cross security levels. For example, the original NT kernel is very good, but it has been slowly compromised by the surrounding software.
It would be possible to make a dedicated system into a trusted platform, for example, an ATM. It is practically very difficult to implement a genera; purpose system in a trusted way.
The guy sold exaclty what he said, a computer RUNNING at a given speed.
Running in relation to automobiles, implies functionality.
If someone sold me a system running at 2.4GHz, I would expect it to be doing that, rather than parked on the metaphorical hard shoulder waiting for the automobile club.
Under EU law, there is the concept of "fitness for the purpose for which it was sold". Most states have this as well. If your co mputer ocasionally glitches whilst playing a game, it isn't a major issue. If it junks your accounts, then it maybe somewhat more important. By seliing a subspec part,, the dealer has made themselves liable.
Video and web conferencing can do a lot of things but you often still need physical bodies to do deals because a lot happens away from the conference room.
Certain British politicians were hopping over to NY on an almost daily basis during the lead-up to the Iraq crisis. They certainly used and needed Concorde with speeches in the Commons in the morning and then speeches in the UN later in the day.
Other Concorde Frequent Flyers include deal-makers from banking and industry. It has certainly helped the City and New York to remain close.
In many European countries we use the prefix to identify the network and the caller pays. If you call from one network to another, you usually pay more than a call within. The prices from landlines to a network also can depend upon the network, especially as the landlines are sometimes provided by a particular operator.
It is possible to use least-cost-routing to find the cheapest call reseller by the dialling code. Unfortunately, if the dialling code can relate to more than one company, then LCR can't work unless you can program in 'exceptions'.
Even if the design was properly prepared, it often isn't by another plumber. A plumber is a specialist who knows about pipes, probably more than an architect or civil engineer. Even when they are being given direction by a 'responsible person' they are still supposed to check for and question mistakes.
It is only in the special case of a plumber working under supervision of another that they can avoid responsibility.
Lets get back to IT. If I work directly for a client, it is clear who has responsibility that my work is correct (ME!!!!). If I work through an agency, then the body shop has no ability to supervise my work other than to check the numbers on timesheets and invoices. If I work as an external through a consultancy company on a client's project, theoretically, the consultancy should take some responsibility for my work as it is presented to the client in their name rather than mine.
These days, the distinction between an agency and a consultancy becomes blurred. Frequently, a consultancy has no permanent staff working on the client's project and they are there only in name.
If I get a plumber to put some pipe-work in and I try to dictate how the work is done, the plumber is perfectly right to expülain to me that, for example, puting hot water through pvc is not a good idea. A plumber is regarded as a contractor but they can still say that your proposed solution is unworkable.
If you have a sign-off on a project then if you aren't happy with something, you are within your rights not to sign. However, you had better have justification, i.e., not liking Microsoft isn't good enough.
As a real consultant, you are paid for your opinion. The opinion you present is not the final decision, that is usually for the business to decide. I agree that you should present not only the recommendation but also your criteria and accept that you do not know the full picture (i.e., Steve Ballmer may sit on their board).
That was a good gesture, but the company should take the view that it should be treated as any other business gift. If a company has a long-standing relationship with a person working at another company, then small gifts may be appropriate under special circumstances, and in the same way, a small gift may be made to a contractor.
For parties, it is a little different. My rule is that if it is a project related event and that contractor is directly and deeply involved with a project then they should be invited. Ironically, this kind of clobbers invitations to the company christmas/new year's party whilst launch parties and team-building events are ok.
The joke that is often repeated is that Disney doesn't want anyone doing to them what they did to the Brothers Grimm and others.
The concept of copyright and patent law is to recognise that ideas should end up in the public domain so that they can be reused. The pool of ideas is a general 'good' and becomes less useful if ithe distribution is limited, that is the person who produces the next idea needs the ability to exploit the first (and that person may not have the $10 mil for a licence fee).
So the government sticks its oar in and allows the creator of a sufficiently novel idea to exclusively benefit for a limited period of time.
The problem is that sufficiently novel and limited period of time seem to be somewhat flexible concepts. The novelty is a particularly odious factor because of the number of court disputes that it leads to.
As you say, we have an artificial market in information and as what is transferred is intangible (I don't buy a CD, I buy the right to listen to that CD privately), the normal producer/consumer relationship associated with physical goods disappears. We have a government licensed monopoly that has a questionable legal basis, over a variable length of time (pay your congress-critter for a new copyright extension) that is easily flouted. Under any basis, I would agree with you and say that there is a deep problem here. If we add the number of persons who aren't recognising thiese laws, then there is something that needs fixing quickly. The alternative of stricter penalties is not the answer.
As the other poster noted, alpha particles are easily blocked. Cosmic Rays arehigh energy particles originating from somewhere in space - not our sun, it isn't powerful enough. These particles tend to hit materials and cause secondary particles.
One of the ceramics used in making earlier memory chips was quite good at turning Cosmic Rays into alpha and beta particles which essentially could cause single bit errors.
No worries, because then people tended to have ECC and it was used frequently, at least in the order of once per montdh or so. The OS was fairly tolerant and if an error could be corrected, the page was logged for exercising later and was usually returned to the active pool. When a page stayed 'bad', then that board would be flagged for replacement. Some systems even called 'home' and specified that a part needed repairing (Remember HAL and the AE35).
Modern memory uses materials that are less susceptible, but it still is a good idea to use ECC on servers. On real-time control systems, I'm a firm believer in multiple architectures, and program implementations. For example, the Airbus uses three independently programmed systems running on Intel and Motorola architectures, so it is unlikely that a bug will occur on all three.
Same rul in the UK and Ireland, I guess we are going with Slashdot's home TZ The thing is some of those posts would be a little early then. In any case, the TV companies would always broadcast theirs in the evening (as in the famous Spaghetti Harvest by the BBC).
Decoding doesn't really seem to be a problem on my slower systems (400MHz and 500MHz). The thing that really helps performance is when the video hardware and the driver supports scaling, for example, using the ATI chips with the GATOS drivers.
I don't know how much RH pay for their bandwidth, I guess it isn't small. Mirrors help womewhat but whether it is a full iso or just a big rpm (say the kernel or glibc), the distribution method has major problems.
Early release of the ISOs isn't what we pay for with RHN, it is priority access to updates. However, I sometimes find that I can fetch them faster with apt-get for rpm. The early release was just a nice little plus. Regrettably, from all reports, their servers are still crawling, even though they are limiting to subscribers. This is one reason why I have always waited a couple of weeks before grabbing a release.
A swarming version of up2date would be really nice (and a good thing for RHN subscribers), it would also help for people with lots of systems under support. Why fetch that kernel from redhat when another local system already has it.
Actually, it is quite old hat, Benefon, the other Finnish mobile phone company has been making them for outdoor freaks and rescue workers. The phone has two modes, one as a straight GPS and the other as a locator using the lat/long info from another so equipped phone to tell you how to get to it.
There is even a variant for hunting dogs, so that hunters can keep track of their dogs through the dense forests.
Re:It depends - and not in investment banking
on
Eclipse 2.1 Released
·
· Score: 1
erl may not be the best but it is ubiquitous. As I mention elsewhere, I prefer other scripting languages such as Python or Ruby. The thing is, that I don't particularly see any reason to code business logic in Java as above other languages and if we are going interpreted, then why not the whole way?
I have already worked on code that is ten to fifteen years old. Believe it or not, Java isn't really transparent enough because you don't want to go into 10-15 year old libraries to find out how a module does things. In the past we did it in Cobol, now we use other languages but on thr backend, Java isn't one of them.
If you have them as freelancers the no problem. If you have registered your business there, you can get caught up with everything from certification of premises through to employment taxes and social security contributions.
To be fair that was a LAN bridged over a lease line. Some queuing but we tried to tune the prioritising based on protocol type and system.
I agree, a noisy phone line is bad - mostly because the error correction doesn't mix well with the PPP or whatever you are throwing up it. A clean line seems to be ok though.
Re:It depends - and not in investment banking
on
Eclipse 2.1 Released
·
· Score: 1
Personally I don't particularly like Perl, but I love CPAN and Perl is in any case still a bit more mainstream than Python or Ruby.
Satellites are wonderful things, a real gods gift for recon. Regrettably, the ones that fly low enough to give the best pictures can't be in geosynchronous orbits. You can only see when the satellite is above your horizon and then it may be over you for just a few minutes and then a gap of half an hour or more before the next pass.
Slashdot Mirror
Personally, I think that the NT kernel seems quite ok. The trouble is that it is closed (so I can't easily understand what is going on) and it is surrounded by layers of crap. I agree that the Linux people could have learned a lot from the synchronisation objects in NT, but they could also have leaned a lot from older systems like VMS, which were extremely well documented. However, better asynch support is coming soon.
However, the next big area for Linux is resource locking (needed for clusters). NT does it, but not very well. VMS did it better (and still does, from last time I used it).
Iain Banks also has Orbitals (Ringworlds) in his Culture universe. The thing is aren't we talking about vehicles, which generally excludes B5, ringworlds and orbitals. A GSV is probably the largests actual moving space ship in current science fiction.
The point is that with TCPA, the hardware checks the signature on the software and the software allows only authorised updates. Great idea, and the principle has been discussed for a long time. The problem is that if there is any duff software in the trusted path then the system can be subverted. Then there is the issue of how to organise keys in a non-hierarchical environment.
They got rid of the VAX and put Alphas as hosts, but still running VMS and the host system behind Eurex is more COBOL than anything else. There is even a little FORTRAN there as well (a file hacking tool). Eurex is the world's biggest electronic derivatives exchange.
Even when my password hits the https client software, how do I know that the information is really being sent securely? I don't.
The counter example used by the digital rights people is that when they send me a key to access controlled media, how can they be certain that I don't intercept the decoded bit stream?
In the first case, it is reasonable to have a trusted platform because the user can choose to accept what software he runs. In particular it can allow me to differentiate between an allowable update and one that isn't.
In the second, then then the owner/user of the system can not be permitted to have control. If the user is permitted to have full control then the platform must disclose to the access granter that the link between the media decryption engine and the output can no longer be trusted.
One can argue that the first is reasonable but the second would prevent anyone from looking at digitally licensed media on an open computing platform such as Linux.
In any case, this all supposes that the platform as installed, is indeed secure. It probably isn't. Even systems that implement a good security reference monitor can be compromised by poor configuration and software layers that cross security levels. For example, the original NT kernel is very good, but it has been slowly compromised by the surrounding software.
It would be possible to make a dedicated system into a trusted platform, for example, an ATM. It is practically very difficult to implement a genera; purpose system in a trusted way.
If someone sold me a system running at 2.4GHz, I would expect it to be doing that, rather than parked on the metaphorical hard shoulder waiting for the automobile club.
Under EU law, there is the concept of "fitness for the purpose for which it was sold". Most states have this as well. If your co mputer ocasionally glitches whilst playing a game, it isn't a major issue. If it junks your accounts, then it maybe somewhat more important. By seliing a subspec part,, the dealer has made themselves liable.
Some people have done this!!!!
Certain British politicians were hopping over to NY on an almost daily basis during the lead-up to the Iraq crisis. They certainly used and needed Concorde with speeches in the Commons in the morning and then speeches in the UN later in the day.
Other Concorde Frequent Flyers include deal-makers from banking and industry. It has certainly helped the City and New York to remain close.
It is possible to use least-cost-routing to find the cheapest call reseller by the dialling code. Unfortunately, if the dialling code can relate to more than one company, then LCR can't work unless you can program in 'exceptions'.
It is only in the special case of a plumber working under supervision of another that they can avoid responsibility.
Lets get back to IT. If I work directly for a client, it is clear who has responsibility that my work is correct (ME!!!!). If I work through an agency, then the body shop has no ability to supervise my work other than to check the numbers on timesheets and invoices. If I work as an external through a consultancy company on a client's project, theoretically, the consultancy should take some responsibility for my work as it is presented to the client in their name rather than mine.
These days, the distinction between an agency and a consultancy becomes blurred. Frequently, a consultancy has no permanent staff working on the client's project and they are there only in name.
If you have a sign-off on a project then if you aren't happy with something, you are within your rights not to sign. However, you had better have justification, i.e., not liking Microsoft isn't good enough.
As a real consultant, you are paid for your opinion. The opinion you present is not the final decision, that is usually for the business to decide. I agree that you should present not only the recommendation but also your criteria and accept that you do not know the full picture (i.e., Steve Ballmer may sit on their board).
For parties, it is a little different. My rule is that if it is a project related event and that contractor is directly and deeply involved with a project then they should be invited. Ironically, this kind of clobbers invitations to the company christmas/new year's party whilst launch parties and team-building events are ok.
I was amused to see that Mao is one of the authors.
The concept of copyright and patent law is to recognise that ideas should end up in the public domain so that they can be reused. The pool of ideas is a general 'good' and becomes less useful if ithe distribution is limited, that is the person who produces the next idea needs the ability to exploit the first (and that person may not have the $10 mil for a licence fee).
So the government sticks its oar in and allows the creator of a sufficiently novel idea to exclusively benefit for a limited period of time.
The problem is that sufficiently novel and limited period of time seem to be somewhat flexible concepts. The novelty is a particularly odious factor because of the number of court disputes that it leads to.
As you say, we have an artificial market in information and as what is transferred is intangible (I don't buy a CD, I buy the right to listen to that CD privately), the normal producer/consumer relationship associated with physical goods disappears. We have a government licensed monopoly that has a questionable legal basis, over a variable length of time (pay your congress-critter for a new copyright extension) that is easily flouted. Under any basis, I would agree with you and say that there is a deep problem here. If we add the number of persons who aren't recognising thiese laws, then there is something that needs fixing quickly. The alternative of stricter penalties is not the answer.
One of the ceramics used in making earlier memory chips was quite good at turning Cosmic Rays into alpha and beta particles which essentially could cause single bit errors.
No worries, because then people tended to have ECC and it was used frequently, at least in the order of once per montdh or so. The OS was fairly tolerant and if an error could be corrected, the page was logged for exercising later and was usually returned to the active pool. When a page stayed 'bad', then that board would be flagged for replacement. Some systems even called 'home' and specified that a part needed repairing (Remember HAL and the AE35).
Modern memory uses materials that are less susceptible, but it still is a good idea to use ECC on servers. On real-time control systems, I'm a firm believer in multiple architectures, and program implementations. For example, the Airbus uses three independently programmed systems running on Intel and Motorola architectures, so it is unlikely that a bug will occur on all three.
In Russia, April Fool's lasts the whole day.
Decoding doesn't really seem to be a problem on my slower systems (400MHz and 500MHz). The thing that really helps performance is when the video hardware and the driver supports scaling, for example, using the ATI chips with the GATOS drivers.
Apple do some great media stuff, but they are not exactly on a budget.
Early release of the ISOs isn't what we pay for with RHN, it is priority access to updates. However, I sometimes find that I can fetch them faster with apt-get for rpm. The early release was just a nice little plus. Regrettably, from all reports, their servers are still crawling, even though they are limiting to subscribers. This is one reason why I have always waited a couple of weeks before grabbing a release.
A swarming version of up2date would be really nice (and a good thing for RHN subscribers), it would also help for people with lots of systems under support. Why fetch that kernel from redhat when another local system already has it.
There is even a variant for hunting dogs, so that hunters can keep track of their dogs through the dense forests.
I have already worked on code that is ten to fifteen years old. Believe it or not, Java isn't really transparent enough because you don't want to go into 10-15 year old libraries to find out how a module does things. In the past we did it in Cobol, now we use other languages but on thr backend, Java isn't one of them.
If you have them as freelancers the no problem. If you have registered your business there, you can get caught up with everything from certification of premises through to employment taxes and social security contributions.
I agree, a noisy phone line is bad - mostly because the error correction doesn't mix well with the PPP or whatever you are throwing up it. A clean line seems to be ok though.
Personally I don't particularly like Perl, but I love CPAN and Perl is in any case still a bit more mainstream than Python or Ruby.
Satellites are wonderful things, a real gods gift for recon. Regrettably, the ones that fly low enough to give the best pictures can't be in geosynchronous orbits. You can only see when the satellite is above your horizon and then it may be over you for just a few minutes and then a gap of half an hour or more before the next pass.