yes. --begin rant. Mainframes died. Why? Tried to turn mainframe into a mess of PCs. Ineffective and costly. Windows NT. This will bring back the mainframes. "Zero Administration" -- BIG LIE. Central Point of control. Another BIG LIE. Conway's law is basic. I don't remember the exact quote, but it goes something like "The structure of a software system is isomorphic to the structure of the organization that produces it." Try Control Panel with a different configuration goodie hidden behind each hieroglyphic. Other configuration in other strange places. Still more only via registry. Microsoft Centralized Control? Hardly. Ken Iverson had control over APL. Linus has control over Linux kernel. Probably a few others, but not many.
I seem to recall that the requirements included a full published audited TPC-D result for the full TPC-D test suite. As for the "results" only, seems like regurgitating precomputed results in about one second could be done with Basic on DOS on an old 486. Take a few days (weeks?) to load it though.
I have to disagree. This is the kernel. Throwing the term "stupid" around is appropriate even if unpleasant. The kernel is a bad place for stupidities. For something to play with, "Cool Ideas" are good. For something that everything else depends on, "Cool Ideas" are not good. If the cost of a good kernel is being labeled stupid for something that is 98% right, the price is cheap. As for the idea of a microkernel, it sounds good, but if I understand Linus, it works out too much like emulating a VAX on x86. It loses more than it gains.
Take a fast look at AutoLisp. It really is a lisp, but it is a baby lisp, with extensions to access some internal AutoCad stuff. Despite being infantile, it adds enormous power and flexibility to AutoCad. AutoLisp is missing basic stuff like DEFEXPR so you cannot make your own SETQ function. Advanced stuff like backquote is way off the horizon. Someone who knows lisp, not I -- completely lost trying to make sense of backquote, can fairly readily and accurately tell what kind and level a lisp is.
This may help explain. a little. Subtle timing differences. Different access patterns. Could go either way. If you have a screwey BIOS, Linux will solve your problems. I would not trust Windows error messages -- often way off.
Very enjoyable read. Linus is very much concerned with the essentials. I think M$ suffers a more fundamental problem. Interfaces. APIs. Big fat bloated interfaces. Linus' "The first very basic rule is to avoid interfaces." is dead on target. You can recover from a bad implementation. You cannot recover from a bad interface. A bad job of doing the right thing works better than a good job of doing the wrong thing. Note: You cannot get rid of all the interfaces, but other things equal, the fewer the better. Much better. Ever wonder WHY Windows requires a reboot after minor configuration changes? M$ has lost control of their own creation. For contrast, I think I've seen something about compiling a new Linux kernel and switching to it without rebooting! A slight pause to switch in-core stuff and back in business. Risky maybe, but you don't even consider such shenanigans unless you are in total control. It is going to be very interesting to watch the development of Linux. I would go so far as to compare/. coverage of Linux et al with CNN's coverage of the Gulf War. Thanks Rob.
... "But the coup de grace is the fact that IBM did this with freely available Beowolf clustering technology and a copy of Red Hat Linux purchased the day before the clustering demo." The point is that IBM just "threw together" a fairly impressive demo. They brought the hardware. They were so "prepared" that somebody had to buy the software the day before. Hehe. Hehe;-) Has anyone tried to do a Beowulf cluster of NT machines? Would you want to demo them the day after tomorrow?
Some of us read/. BECAUSE of the MS bashing. Sometimes it's even clever and original. What IS pathetic is that I have to use the (MS) stuff. $500 for 1/24/40/whatever copies of Bill Gates' whatever is just too good a target to leave alone.
yes.
--begin rant.
Mainframes died. Why? Tried to turn mainframe into a mess of PCs. Ineffective and costly.
Windows NT. This will bring back the mainframes. "Zero Administration" -- BIG LIE.
Central Point of control. Another BIG LIE. Conway's law is basic. I don't remember the exact quote, but it goes something like "The structure of a software system is isomorphic to the structure of the organization that produces it." Try Control Panel with a different configuration goodie hidden behind each hieroglyphic. Other configuration in other strange places. Still more only via registry. Microsoft Centralized Control? Hardly. Ken Iverson had control over APL. Linus has control over Linux kernel. Probably a few others, but not many.
I seem to recall that the requirements included a full published audited TPC-D result for the full TPC-D test suite. As for the "results" only, seems like regurgitating precomputed results in about one second could be done with Basic on DOS on an old 486. Take a few days (weeks?) to load it though.
I have to disagree. This is the kernel. Throwing the term "stupid" around is appropriate even if unpleasant. The kernel is a bad place for stupidities. For something to play with, "Cool Ideas" are good. For something that everything else depends on, "Cool Ideas" are not good. If the cost of a good kernel is being labeled stupid for something that is 98% right, the price is cheap.
As for the idea of a microkernel, it sounds good, but if I understand Linus, it works out too much like emulating a VAX on x86. It loses more than it gains.
Take a fast look at AutoLisp.
It really is a lisp, but it is a baby lisp, with extensions to access some internal AutoCad stuff. Despite being infantile, it adds enormous power and flexibility to AutoCad.
AutoLisp is missing basic stuff like DEFEXPR so you cannot make your own SETQ function. Advanced stuff like backquote is way off the horizon.
Someone who knows lisp, not I -- completely lost trying to make sense of backquote, can fairly readily and accurately tell what kind and level a lisp is.
This may help explain. a little.
Subtle timing differences. Different access patterns. Could go either way.
If you have a screwey BIOS, Linux will solve your problems. I would not trust Windows error messages -- often way off.
Very enjoyable read. Linus is very much concerned with the essentials. /. coverage of Linux et al with CNN's coverage of the Gulf War. Thanks Rob.
I think M$ suffers a more fundamental problem. Interfaces. APIs. Big fat bloated interfaces.
Linus' "The first very basic rule is to avoid interfaces." is dead on target. You can recover from a bad implementation. You cannot recover from a bad interface. A bad job of doing the right thing works better than a good job of doing the wrong thing. Note: You cannot get rid of all the interfaces, but other things equal, the fewer the better. Much better.
Ever wonder WHY Windows requires a reboot after minor configuration changes? M$ has lost control of their own creation. For contrast, I think I've seen something about compiling a new Linux kernel and switching to it without rebooting! A slight pause to switch in-core stuff and back in business. Risky maybe, but you don't even consider such shenanigans unless you are in total control.
It is going to be very interesting to watch the development of Linux. I would go so far as to compare
... "But the coup de grace is the fact that IBM did this with freely available Beowolf clustering technology and a copy of Red Hat Linux purchased the day before the clustering demo." ;-) Has anyone tried to do a Beowulf cluster of NT machines? Would you want to demo them the day after tomorrow?
The point is that IBM just "threw together" a fairly impressive demo. They brought the hardware. They were so "prepared" that somebody had to buy the software the day before. Hehe. Hehe
Yep. Imagine the president of General Motors driving a Lincoln.
Some of us read /. BECAUSE of the MS bashing. Sometimes it's even clever and original. What IS pathetic is that I have to use the (MS) stuff.
$500 for 1/24/40/whatever copies of Bill Gates' whatever is just too good a target to leave alone.