>Basically if you have an un-parallelizable part of the problem, then even with infinite numbers of parallel processors you can't get better than the time to solve that un-parallelizable part. Mainframes (and for that matter supercomputers) dominate where the problem can't be broken down into parallel chunks (for example you really want a bank to keep its database of accounts current so that it doesn't allow someone to withdraw the same money twice). Thanks for posting. This is basic and easy to miss among the hype and latest craze d'jour. Using a mainframe to emulate a mess of PCs is an expensive misuse of mainframe. Trying to emulate a mainframe with a mess of PCs gets very complicated and expensive for something that doesn't quite work.
S/390 instruction set is very different from x86. Disclaimer: I haven't seen S/390 Principals of Operation, but can safely assume it is consistent with the old 360 architecture with extensions to cross the 16 Meg barrier. S/390 is big endian, x86 is little endian. S/390 is EBCDIC, x86 is ASCII. S/390 uses different op-codes for signed and unsigned binary, x86 uses same op-codes and distinguishes via use of condition codes.
What make you think it is being sold back to the open source community? Anyone with a clue would set up their own BSD box and control and admin it themselves. IBM does not control BSD and is even less likely to control BSD than to control Linux. IBM among others supports Apache. IBM does not control Apache. As to BSD's technical superiority, the aproach and philosophy are different. Saying BSD is technically superior is not quite the same as saying Linux is technically inferior. Anything much better in one will soon show up in the other anyway. Open source is much stronger with BOTH! Why do I get the impression that I am responding to a MS troll;-)
It's for the benefit of the clueless. Development kernels are probably not the place for newbies to start. I thought it was a tad funny myself;-)
Re:Freshmeat whiners, let me see if I get this
on
Linux 2.3.48 Released
·
· Score: 1
2.4 will be out before long. Posting about the latest 2.3.xx gives a preview of what to expect. It also has to help to get some widespread testing of the later development kernels. Agreed, I suspect many of the whiners are Micros~1 trolls. Just think, the latest development kernel *is* more newsworthy that the official launch of w2k.
Why do you think IBM has been supporting Linux? Look a few years down the road. The computers get bigger, faster, more complicated. There are more ways to get it wrong, and the interactions get more subtle. Hackers handle a part of the problem space that IBM and big iron cannot. With journaling file systems now under GPL, how long before hackers with surplus 486 boxen and bad disks try to see what it takes to fubar a system, and then fix things so it can't happen in the future. I cannot imagine IBM doing that in a closed, secret test facility. Wrong image for the company. I get the impression that the influx of capital into the Linux companies is being used to shore up the infrastructure and cure some of the grungy problems that require paid help.
Is there money to be made from Linux? Consider if the software you are using were GPL'd (free speech) and downloadable free (beer), would your company drop the HP support? I didn't think so.;-)
IIRC, in early IBM MVS (360 vintage), memory was allocatable from several subpools and subpools could be allocated/freed themselves. Surely 30+ years of "progress" has not made storage allocation more primitive.
In reality you would want to allocate a "few" large blocks from the OS, then sub-allocate pieces of those blocks within a subsystem. When you have shared memory and long-lived processes, things get interesting.
Try one of the BSDs, say FreeBSD. It will at least give you a different perspective on things unix. You might even like it. Warning, pc "partitions" are BSD "slices" which are subdivided into BSD "partitions". Find an editor you can use. (Hint: Emacs uses Ctrl-x Ctrl-c to exit). I agree that in another year, Linux should have its act together enough to be useable by us ordinary mortals.
Very true, but the "down with Microsoft!" is not coming from Linux (or *BSD) users, but from us Microsoft users who are looking for a way out.
For evidence. There is a lot of good technical information about Microsoft software hidden in the threads (but never from the Microsoft astroturfers). Microsoft bash fests seem to draw an inordinate amount of interest, but there is a strong impression that the heavy Linux/BSD/Unix users consider it a waste of bandwidth.
>>Name any engineering discipline, and I'll tell you the sciences that predated it.
Roman viaducts and aqueducts? I'm not sure how good the ancient Romans were at science.
Between theory and practice, sometimes theory leads, usually practice leads, but there is a thin line between them which cannot be broken.
Laplace transforms have been used by engineers for a long time. Adequate mathematical theory of such _may_ exist now.
From Chapter 3, Random Numbers: "Anyone who considers arithmetical methods of producing random digits is, of course, in a state of sin. -- John Von Neumann (1951) "Round numbers are always false." -- Samuel Johnson (c. 1750)
However, it seems like the world at large is beginning to realize they have been taken by a scam artist. Just wait till Win00 gets put to production stress. My gut feel is that the beta testers have played with it a bit, but have stayed far, far away from actually stressing it.
"However, wholesale migration to Windows NT Server has not yet been attempted." Microsoft's version of "The Truth". There's more "Truth in Advertising" in a carnival side show.
I think he's getting at the dichotomy between "hacker Linux" and "commercial Linux". The difference is more in the buyer that what is bought, say like the difference between a Cheapbyte's RedHat CD and an expensive supported RedHat. If you wonder why a CxO (or PHB) would buy a $1,000-range RedHat when he could download it for free, image he has a problem, and direct him to Ask Slashdot. He is not interested in flame wars or Micros~1 bashing, and would panic at a hint of the idea. I have no real idea as to what the solution will be, but consider that symbiosis can be defined as mutual parasitism. Look for RedHat, IBM, etc. to profit (not profiteer) from Linux. Look for Linux (and/or *BSD) to be THE Enterprise OS in about 3 to 5 years.
Bootstrap from DOS. Novell, at least for some versions, does exactly that. For Win95 and probably Win98, you can intall real-mode drivers via config.sys which are used by Win95/98. These are DOS drivers, not Windows drivers. For them to work, they *must* be talking to DOS, not to Windows. Seems like you can run Win95 under Quarterdecks QEMM. While I'm at it, what's up with the DOS2 limit of 20 file handles, 5 of which are taken up by StdIn, StdOut, StdErr, PRN, and AUX?
It's nothing like "All the news that's fit to print". It's more like finding interesting rocks and seeing what crawls out from under. Slashdot's coverage of "news" that does not matter is poor and I like it like that. I think their coverage of stuff that does matter is quite good and is mostly in the threads.
Slashdot Mirror
evil empire
#1
>Basically if you have an un-parallelizable part of the problem, then even with infinite numbers of parallel processors you can't get better than the time to solve that un-parallelizable part. Mainframes (and for that matter supercomputers) dominate where the problem can't be broken down into parallel chunks (for example you really want a bank to keep its database of accounts current so that it doesn't allow someone to withdraw the same money twice).
Thanks for posting. This is basic and easy to miss among the hype and latest craze d'jour.
Using a mainframe to emulate a mess of PCs is an expensive misuse of mainframe. Trying to emulate a mainframe with a mess of PCs gets very complicated and expensive for something that doesn't quite work.
Linux: more like a tricycle to a supertanker.
S/390 instruction set is very different from x86.
Disclaimer: I haven't seen S/390 Principals of Operation, but can safely assume it is consistent with the old 360 architecture with extensions to cross the 16 Meg barrier.
S/390 is big endian, x86 is little endian.
S/390 is EBCDIC, x86 is ASCII.
S/390 uses different op-codes for signed and unsigned binary, x86 uses same op-codes and distinguishes via use of condition codes.
Silly question. Just *how* would you gateway between TCP/IP and NetBEUI?
What make you think it is being sold back to the open source community? Anyone with a clue would set up their own BSD box and control and admin it themselves. ;-)
IBM does not control BSD and is even less likely to control BSD than to control Linux.
IBM among others supports Apache. IBM does not control Apache.
As to BSD's technical superiority, the aproach and philosophy are different. Saying BSD is technically superior is not quite the same as saying Linux is technically inferior. Anything much better in one will soon show up in the other anyway. Open source is much stronger with BOTH!
Why do I get the impression that I am responding to a MS troll
It's for the benefit of the clueless. Development kernels are probably not the place for newbies to start. I thought it was a tad funny myself ;-)
2.4 will be out before long. Posting about the latest 2.3.xx gives a preview of what to expect. It also has to help to get some widespread testing of the later development kernels. Agreed, I suspect many of the whiners are Micros~1 trolls.
Just think, the latest development kernel *is* more newsworthy that the official launch of w2k.
Why do you think IBM has been supporting Linux? Look a few years down the road. The computers get bigger, faster, more complicated. There are more ways to get it wrong, and the interactions get more subtle. Hackers handle a part of the problem space that IBM and big iron cannot. With journaling file systems now under GPL, how long before hackers with surplus 486 boxen and bad disks try to see what it takes to fubar a system, and then fix things so it can't happen in the future. I cannot imagine IBM doing that in a closed, secret test facility. Wrong image for the company. I get the impression that the influx of capital into the Linux companies is being used to shore up the infrastructure and cure some of the grungy problems that require paid help.
;-)
Is there money to be made from Linux? Consider if the software you are using were GPL'd (free speech) and downloadable free (beer), would your company drop the HP support? I didn't think so.
The devil is in the details.
IIRC, in early IBM MVS (360 vintage), memory was allocatable from several subpools and subpools could be allocated/freed themselves. Surely 30+ years of "progress" has not made storage allocation more primitive.
In reality you would want to allocate a "few" large blocks from the OS, then sub-allocate pieces of those blocks within a subsystem. When you have shared memory and long-lived processes, things get interesting.
If you are setting the record straight, let's not have 2 atoms of oxygen vanish.
2H2 + 2O2 ---> 2H2O + 1O2
Try one of the BSDs, say FreeBSD. It will at least give you a different perspective on things unix. You might even like it. Warning, pc "partitions" are BSD "slices" which are subdivided into BSD "partitions".
Find an editor you can use. (Hint: Emacs uses Ctrl-x Ctrl-c to exit).
I agree that in another year, Linux should have its act together enough to be useable by us ordinary mortals.
An occasional voice of reason is very welcome. /. Enjoy.
For myself, Welcome to
Very true, but the "down with Microsoft!" is not coming from Linux (or *BSD) users, but from us Microsoft users who are looking for a way out.
For evidence. There is a lot of good technical information about Microsoft software hidden in the threads (but never from the Microsoft astroturfers). Microsoft bash fests seem to draw an inordinate amount of interest, but there is a strong impression that the heavy Linux/BSD/Unix users consider it a waste of bandwidth.
>>Name any engineering discipline, and I'll tell you the sciences that predated it.
Roman viaducts and aqueducts? I'm not sure how good the ancient Romans were at science.
Between theory and practice, sometimes theory leads, usually practice leads, but there is a thin line between them which cannot be broken.
Laplace transforms have been used by engineers for a long time. Adequate mathematical theory of such _may_ exist now.
From Chapter 3, Random Numbers:
"Anyone who considers arithmetical methods of producing random digits is, of course, in a state of sin.
-- John Von Neumann (1951)
"Round numbers are always false."
-- Samuel Johnson (c. 1750)
"an 8.5x11 word 2000 document"
Look, barf, bury it.
Not much delay.
If W2K did _NOT_ launch (with 64k+ bugs), it would be news.
Well, it has so far. Look at Bill's billions.
However, it seems like the world at large is beginning to realize they have been taken by a scam artist. Just wait till Win00 gets put to production stress. My gut feel is that the beta testers have played with it a bit, but have stayed far, far away from actually stressing it.
"However, wholesale migration to Windows NT Server has not yet been attempted."
Microsoft's version of "The Truth".
There's more "Truth in Advertising" in a carnival side show.
I think he's getting at the dichotomy between "hacker Linux" and "commercial Linux". The difference is more in the buyer that what is bought, say like the difference between a Cheapbyte's RedHat CD and an expensive supported RedHat. If you wonder why a CxO (or PHB) would buy a $1,000-range RedHat when he could download it for free, image he has a problem, and direct him to Ask Slashdot. He is not interested in flame wars or Micros~1 bashing, and would panic at a hint of the idea.
I have no real idea as to what the solution will be, but consider that symbiosis can be defined as mutual parasitism. Look for RedHat, IBM, etc. to profit (not profiteer) from Linux. Look for Linux (and/or *BSD) to be THE Enterprise OS in about 3 to 5 years.
...the basic point and click stage that is required for really good NT administration...
That parses as really good NT administration requires [only] the basic point and click stage.
Bootstrap from DOS.
Novell, at least for some versions, does exactly that.
For Win95 and probably Win98, you can intall real-mode drivers via config.sys which are used by Win95/98. These are DOS drivers, not Windows drivers. For them to work, they *must* be talking to DOS, not to Windows. Seems like you can run Win95 under Quarterdecks QEMM.
While I'm at it, what's up with the DOS2 limit of 20 file handles, 5 of which are taken up by StdIn, StdOut, StdErr, PRN, and AUX?
Strong agreement.
It's nothing like "All the news that's fit to print". It's more like finding interesting rocks and seeing what crawls out from under. Slashdot's coverage of "news" that does not matter is poor and I like it like that. I think their coverage of stuff that does matter is quite good and is mostly in the threads.
Run NASA on W2k and launch Bill Gates into space.