The issue with technology and the govt. isn't Clinton per se, since he just happened to be around at the time when technology was reaching everyone on the planet. Govt. has never been known for its speed. The Constition is built to make things hard to do for a reason. Thus, govt. (be it republican or democrat) will take a while to adjust to new technology and new ideas.
Actually, this particular law leaves a lot of discretion in the hands of the Executive Branch, which could change the rules at the drop of a hat (but doesn't). Part of the problem is the way policies are made in government (insert tongue in cheek (but only partway) and feel free to correct me if I miss a step):
Generally, the way it works is: Official A (in this case, the President, but it could be the Secretary of Commerce or virtually any PHB in government) asks for an expert opinion from an outsider, and gets a document. He asks his staff who needs to see it, then gets those cabinet members to review it. About 80% the time, the outsider has a good idea, the other 20% we start with a fragrant crock.
Official B (in this case, the Attorney General or the FBI director) is still too high on the toem pole to understand the document, so it's referred down to career manager C, who notices that implementing the suggestion would a) reduce their budget, b) reduce their leverage, c) make their job harder. They instruct their staff to "Shoot this down". Staff comes up with a reasonable critique, which is returned to them on a silver platter with a self-destructing sticky note saying "Not strong enough -- use a trowel". If the review was favorable, they receive the platter in a "meeting", where choice phrases are offered as a side dish. Staff gives up on their principles and takes the platter down to the horse stables to load it up. Said platter, now steaming, is returned to B, who returns it to Official A.
A now has a choice: offend an expert that no-one has heard of, or make the papers for overriding B, who he's on record touting as the greatest expert in living memory on the subject for which B was hired. A had to make that kind of recommendation to get B past the review board, so A is stuck with the steaming plate, unless he passes the whole mess to Staffer D, (say, the National Security Advisor) who uses a similar method to return a steaming, loaded plate of different opinion to the big cheese (A).
Now A is really in a bind, and has to propose a policy that averages the agendas of his staffers with the expert opinion. This sometimes stops the stupid suggestions (but remember, it's an average) and it also tends to prevent really smart stuff from going through (almost perfectly, when the current policy isn't similar to the smart policy). This Rube Goldberg method is used to formulate any policy that wasn't in the campaign (and most of the ones that were) by almost every elected official (unless they actually know something about the issue). It also ensures that unless A or B understands the issue, a change of administration does not mean a change of policy (unless, as mentioned above, it was a plank, or it's been a real embarassment and come to someone's attention).
Notice that most of your high officials these days are lawyers, not geeks, so you'll never see a change in the rules on highly technical issues until the rules come to someone's attention by being inconvenient for a big contributor, or until we elect someone whose advisors talk them into making a campaign promise about the problem.
So go out and become a high-priced campaign consultant (the more you charge the better they'll listen), or write me in for President in 2000!
The ASC's arguments that "experts" in intelligence should dictate crypto policy is ludicrous. [...] I simply can't believe that people THIS STUPID are making decisions more important than what they will have for lunch today.
Everything's OK! In the House cafeteria, trained experts are deciding what these people will have for lunch today. (when I visited it was tuna salad on white bread)
We run multiple instances and versions on several NT development servers. I agree it requires tuning to get 'em running right on production servers (where you want to have all possible resources used) but it's possible.
I agree with your conclusion, but some friendly advice (that I will not guarantee, YMMV):
1) If you're using NT4 SP3 or SP4 you may have some NTFS stability problems on large volumes under high load. SP5 apparently fixes that. A major clue is if your lockup results in a file system check. We had problems with 7.x too, but they went away with the 8.05 upgrade + the sp4 patch to the filesystem.
2) Be sure to make at least one mirrored drive for your redo logs (and get your swap space off the RAID too). RAID5 is not as good as mirrored stripe sets (RAID 0+1) for data, but it's doable if you get the logs on a mirrored volume (and you're doing hardware RAID).
3) You're not using virtual RAM, are you? ORACLE will be a serious dog in NT if you're doing any swapping, and it's way too easy to make it eat too much RAM.
1) Why emulate a broken product? Just program for Sybase (or Oracle, or...)
2) Your example for building a "locks" table won't work on a database that doesn't do locking (like SQL server sometimes fails to do). Occasionally two people will attempt to lock at the same time and since the database doesn't lock properly, they'll both succeed.
3) The API solution is the only solution you present that will work, but again, it has to work for the API as well as the database. Implementation of a proper locking mutex is a solved problem, but its implementation is non-trivial in some environments. I remember Amiga (through the 1500) had a prohibition against using Test-and-set instructions because it threw off the time-multiplexing for memory access, for instance.
"You could cut Linux some slack if it were sharply lower in cost per transaction than NT, but that's not the case."
Hee hee hee... at work we're paying Big Bucks for an NT guru to get our email servers working... If I were using Linux we'd be done, it would work and Redhat could support us for... well less than we'll be paying for 3 mos. of NT support...
So is MS giving away stuff to lower their TCO? (cash would work) Apparently I missed that memo...
Slashdot Mirror
everyone on the planet. Govt. has never been known for its speed. The Constition is built to make things hard to do for a reason. Thus, govt. (be
it republican or democrat) will take a while to adjust to new technology and new ideas.
Actually, this particular law leaves a lot of discretion in the hands of the Executive Branch, which could change the rules at the drop of a hat (but doesn't). Part of the problem is the way policies are made in government (insert tongue in cheek (but only partway) and feel free to correct me if I miss a step):
Generally, the way it works is: Official A (in this case, the President, but it could be the Secretary of Commerce or virtually any PHB in government) asks for an expert opinion from an outsider, and gets a document. He asks his staff who needs to see it, then gets those cabinet members to review it. About 80% the time, the outsider has a good idea, the other 20% we start with a fragrant crock.
Official B (in this case, the Attorney General or the FBI director) is still too high on the toem pole to understand the document, so it's referred down to career manager C, who notices that implementing the suggestion would a) reduce their budget, b) reduce their leverage, c) make their job harder. They instruct their staff to "Shoot this down". Staff comes up with a reasonable critique, which is returned to them on a silver platter with a self-destructing sticky note saying "Not strong enough -- use a trowel". If the review was favorable, they receive the platter in a "meeting", where choice phrases are offered as a side dish. Staff gives up on their principles and takes the platter down to the horse stables to load it up. Said platter, now steaming, is returned to B, who returns it to Official A.
A now has a choice: offend an expert that no-one has heard of, or make the papers for overriding B, who he's on record touting as the greatest expert in living memory on the subject for which B was hired. A had to make that kind of recommendation to get B past the review board, so A is stuck with the steaming plate, unless he passes the whole mess to Staffer D, (say, the National Security Advisor) who uses a similar method to return a steaming, loaded plate of different opinion to the big cheese (A).
Now A is really in a bind, and has to propose a policy that averages the agendas of his staffers with the expert opinion. This sometimes stops the stupid suggestions (but remember, it's an average) and it also tends to prevent really smart stuff from going through (almost perfectly, when the current policy isn't similar to the smart policy). This Rube Goldberg method is used to formulate any policy that wasn't in the campaign (and most of the ones that were) by almost every elected official (unless they actually know something about the issue). It also ensures that unless A or B understands the issue, a change of administration does not mean a change of policy (unless, as mentioned above, it was a plank, or it's been a real embarassment and come to someone's attention).
Notice that most of your high officials these days are lawyers, not geeks, so you'll never see a change in the rules on highly technical issues until the rules come to someone's attention by being inconvenient for a big contributor, or until we elect someone whose advisors talk them into making a campaign promise about the problem.
So go out and become a high-priced campaign consultant (the more you charge the better they'll listen), or write me in for President in 2000!
The ASC's arguments that "experts" in intelligence should dictate crypto policy is ludicrous. [...]
I simply can't believe that people THIS STUPID are making decisions more important than what they will
have for lunch today.
Everything's OK! In the House cafeteria, trained experts are deciding what these people will have for lunch today. (when I visited it was tuna salad on white bread)
We run multiple instances and versions on several NT development servers. I agree it requires tuning to get 'em running right on production servers (where you want to have all possible resources used) but it's possible.
I agree with your conclusion, but some friendly advice (that I will not guarantee, YMMV):
1) If you're using NT4 SP3 or SP4 you may have some NTFS stability problems on large volumes under high load. SP5 apparently fixes that. A major clue is if your lockup results in a file system check. We had problems with 7.x too, but they went away with the 8.05 upgrade + the sp4 patch to the filesystem.
2) Be sure to make at least one mirrored drive for your redo logs (and get your swap space off the RAID too). RAID5 is not as good as mirrored stripe sets (RAID 0+1) for data, but it's doable if you get the logs on a mirrored volume (and you're doing hardware RAID).
3) You're not using virtual RAM, are you? ORACLE will be a serious dog in NT if you're doing any swapping, and it's way too easy to make it eat too much RAM.
1) Why emulate a broken product? Just program for Sybase (or Oracle, or...)
2) Your example for building a "locks" table won't work on a database that doesn't do locking (like SQL server sometimes fails to do). Occasionally two people will attempt to lock at the same time and since the database doesn't lock properly, they'll both succeed.
3) The API solution is the only solution you present that will work, but again, it has to work for the API as well as the database. Implementation of a proper locking mutex is a solved problem, but its implementation is non-trivial in some environments. I remember Amiga (through the 1500) had a prohibition against using Test-and-set instructions because it threw off the time-multiplexing for memory access, for instance.
Yeah, but if Huston turned off the astronauts' radio...
I'm in Raleigh -- anyone know of choices here?
:-).
My Granddad also owns some BS stock -- when's the next stockholder's meeting? Can you say "proxy?"
"You could cut Linux some slack if it were sharply lower in cost per transaction than NT, but that's not the case."
... well less than we'll be paying for 3 mos. of NT support...
Hee hee hee... at work we're paying Big Bucks for an NT guru to get our email servers working... If I were using Linux we'd be done, it would work and Redhat could support us for
So is MS giving away stuff to lower their TCO? (cash would work) Apparently I missed that memo...
Cool story -- they didn't make any serious errors, told me something I didn't know, and we even got to see Linus 2.0 :-)