Let's see, there are really only two alternatives:
1)The program allocates some memory, uses it, and then immediately deallocates it. Since the program might not need it for very long (possibly an object-oriented system like you mentioned), this would fall under the "rapidly allocates and deallocates" category.
OR
2)The program allocates some memory, and then waits until it exits (or some other, later time) to deallocate it.
As far as the first line in my post goes, I was trying to imply that any properly written program would be affected at least somewhat by this bug, since only bad programmers will rapidly allocate memory without also (potentially rapidly) deallocating it right after using it.
I think you need to reread my post. Nowhere did I say that programs that allocated a block and then didn't deallocate it until a good while later would be affected. Rather, I said just the opposite.
A program that frequently allocates memory but doesn't deallocate it until much later is either broken, poorly written, or (most likely) both.
I think that this will hit programs that try to be at least somewhat memory efficient the hardest, because those programs will deallocate a block of memory as soon as they're done using it instead of waiting until much later to do it, leaving room for a potential slowdown.
Re:"even bigger than some bacteria"
on
World's Largest Virus
·
· Score: 2, Informative
That's how all viruses reproduce. They latch onto a host cell and inject their own DNA into it. The virus DNA then proceeds to hijack the cell's reproductive mechanisms, forcing it to make more viruses inside itself until the cell becomes so full of these new viruses that it bursts open and dies, thereby unleashing the newly made viruses.
From what I've gathered, Steve Jobs was at the Intel conference representing Pixar, not Apple. As you may remember, Pixar recently began to drop their Sun-powered render farm in favor of a cheaper Intel-powered one. This had nothing to do with Macs vs PCs, Motorola vs Intel, etc.
Similarly, while Apple and Motorola certainly haven't been getting along for a good while now, it doesn't make sense for Apple to switch to a CPU with an entirely new instruction set. Regardless of whether or not OS X runs on x86, all of the Mac OS X software would have to be ported.
What's more, Apple would lose a lot of their customer base, because there's a certain air of eliteness that comes with using a Mac (or, at least there is in the minds of some Mac users;) ), and those customers would drop Apple if they thought it was "just another PC."
And yet more still, it would be like SGI's Intel boxes: nobody wants to pay through the nose for x86 (c'mon SGI, $30 thousand for an Intel box? No thanks), especially when they can get it for cheaper by ordering it from Dell, HP, or building it themselves.
No, I think Dvorak is just being his usual idiot self. Besides, hasn't Apple already announced (maybe not officially) that they're going with IBM's 970 PPC processor? That would certainly make more sense, since Mac developers wouldn't have to port their software to a whole new architecture, and would only require a new motherboard and some small changes to the OS to handle the 64-bit pointers and such.
The black hole of the story, with billions of solar masses, is probably less dense than water...
Actually, no. From what I understand, the "size" of the black hole is actually the size of the event horizon, since the black hole itself has a size of zero and infinite density.
I'd be willing to bet that the Sci-Fi Channel execs had a hand in that ending.
*SPOILERS BELOW* By killing off the two main characters, they can say, "Well, see, we can't continue the show, because Crichton and Arynn are dead." And then either a)the producers snuck in the "To Be Continued" so that the fan outrage over not getting to find out if they really are dead would get the show another season/movie/miniseries to resolve it, or that b)the execs at the Sci-Fi Channel were hedging their bets in case the fans were somehow able to resurrect the show.
It is my understanding that black holes emit a very small amount of radiation, and do it very slowly. Therefore, the black hole would have to be very small (thereby having very little gravity) for this to take place.
Besides, a black hole would have to swallow all the matter within its reach before it could shrink to the size necessary for the effect you describe to take place, because the more matter it takes in, the bigger it gets. And since it would have already taken in all the matter it can get ahold of, the effect you mention wouldn't happen.
Well, funny posts aside, as scales get more and more accurate, we find that the old ways of measuring how much something is (such as the mass of one kilogram) aren't accurate enough (possibly a small amount of variance that the old instruments couldn't measure, but the newer, more precise scales can).
So, we need new, more precise methods of defining things like how much mass one kilogram has.
The headline talks about "weighing" the black hole, so I used the word weight. But I knew I'd get reamed if I didn't mention that it's actually the mass of the black hole, not its weight, so I put that in there too. So much for hedging my bets.
But you're right. I should've written mass and left it at that.
Technically, FAT32 didn't resemble a 32-bit filesystem either; that came when fdisk 32, which actually enabled the 32-bit goodness at the partition level, was released with Win98 OEM version D.
All fdisk can do is make and delete partitions. The reason you needed the fdisk that came with Windows 95 OSR2 or Windows 98 to create FAT32 partitions was because MS didn't give it the ability to create partitions labelled as using the FAT32 filesystem until then. But that is the end of fdisk's involvement.
FAT32 was, in fact, a 32-bit filesystem. It stored things like disk block numbers, etc., in 32-bits. Microsoft's little secret, though, was that only 22 of those 32 bits were actually used. The other 10 were empty, or something like that.
You did not recall correctly. It's 20-bit. 5 nibbles. And yes, the segmentation was horrible.
Then why does my old 286 have 2 megabytes of RAM? 20 bits of address space means that you can access only 1 megabyte at most, which is what the 8086, 8088, and 186 had. Therefor, the 286 would have HAD to have more that 20 address lines (although, as with modern Intel processors, only 20 of the address lines are available when running in real-mode).
2)I'd rather have an OS without the bugs and quirks, thanks. But most programs still ran just fine on DR-DOS, so the point is moot. That, and I was too poor to have very many games, so I didn't really notice.
Win 3 supported every feature of the 386 processor. It could run 32 bit code (although most of the code was 16 bit for compatibility). It could run DOS programs in V86 mode. It supported 4Gb of RAM. That's pretty much every 386 feature accounted for.
You could only run 32-bit programs with the Win32S (yes, there's an S on there) addon. Most programs didn't use it.
4GB of RAM? Really, that's quite astonishing, considering that not even Windows 95 supported that much.
DOS programs in V86 mode? I'll give you that one. Keep in mind that I said Windows 3.1 supported some of the 386's features.
The article claims that DOS tasks where pre-emptively multitasked. This is correct. I thought it was true for 2.0/386 as well, though, but I'm not certain, having never actually used that (I only ever used 2.0 on a 186).
I never stated otherwise. But for all the non-DOS programs, the versions of Windows prior to Windows 95 (and excepting Windows NT) used cooperative multitasking.
1)I didn't say otherwise. You, however, (correctly) state later in your own post that Win95 did in fact thunk to 16-bit code in some places. 2)Again, where did I say otherwise? 3)Where do we disagree on this point?
And MS didn't have anything to do with DOS until version 6.2 or whatever.
Not even close. When MS heard that IBM was designing the PC, and looking for an OS for it, they bought a dinky little OS that nobody had ever heard of called QDOS (Quick and Dirty Operating System), which was itself an x86 hack/clone of the popular CP/M operating system, and made some changes to it so that it would run on a PC.
IBM had been hoping to use actual CP/M, but the company that made it, Intergalactic Digital Research (the Intergalactic part was eventually dropped), was too slow off the mark in doing the x86 port (CP/M was originally for the Z80 microprocessor), so they picked Microsoft with their "new" "Disk Operating System". IBM called it PC-DOS, and shipped it themselves for a while, but all development beyond QDOS was done by Microsoft. MS, of course, also licensed DOS to companies making PC clones, such as Compaq, and eventually (at something like DOS 3.0) started selling it to the public themselves.
Intergalactic Digital Research eventually finished the x86 PC port of CP/M (in something like 1982), but by then it was too late. They tried to pull a reverse-Microsoft on Microsoft by marketing their own version of DOS called DR-DOS (they had dropped the Intergalactic part of their name), but Microsoft had released Windows by then, and were able to out-Microsoft Digital Research by making Windows 3.1 only work on MS-DOS and spreading FUD that DR-DOS wasn't as good as MS-DOS (quite the opposite in reality, though; DR-DOS was vastly more stable than MS-DOS, and the only reason that Microsoft even thought about developing MS-DOS beyond version 3.0 was to compete with Digital Research; their marketing department announced MS-DOS 4.0 before any of the design people had even considered it, and later did the same with MS-DOS 5.0).
1)From the article: "Many longtime PC users trace Windows to the 1990 release of Windows 3.0, the first widely popular version of Windows and the first version of Windows many PC users ever tried. But Microsoft actually released the first version of Windows six years earlier, in 1985." You'd think that Microsoft of all companies could afford to hire somebody who can do basic arithmetic.
2)The text in the article seems to have been taken almost verbatim from the microsoft.com link you posted. No wonder the entire article seemed like it had been written by a marketroid.
DOS was 16-bit real-mode. Windows 95 was 32-bit protected mode, with some 16-bit DOS underpinnings for backwards compatibility, and so that a complete rewrite wouldn't be necessary.
Of course, if DOS had been a decent operating system this wouldn't have been so bad, but DOS was a port of a knock-off of an OS designed for computers with only 64K of RAM. It should have died the minute the 286 came out, or at least when the 386 came out.
I didn't say that Windows 3.1 didn't support the 386, just that it didn't fully support it (meaning no 32-bit protected mode, no 32-bit apps without the special Win32S addon, etc.), contrary to what the article says.
You have to preemptively multi-task a DOS program running in virtual real-mode if you want to run more than one of them, since DOS programs don't have any cooperative multitasking capabilities (cooperative multitasking is basically a system where the processes have to tell the OS, "Ok, go ahead and let another process have a turn now"; preemptive multitasking is where the OS sets a timer (say, for 20 milliseconds), and when this timer is up a specific interrupt is generated, and this interrupt is the kernel's process scheduler, which then resets the timer and switches to a different process, in other words the OS decides when it's time to switch processes, not the processes themselves). OTOH, regular Windows 3.1 programs were still cooperatively multi-tasked, which is retarded and can be annoying for the users (ask anybody who's ever tried to copy a 100 megabyte file on an OS with cooperative multitasking).
Namely that versions of Windows before Win95 didn't fully support the 386 (dunno 'bout NT, never used it), despite what the article claims, still had worthless (and error-prone) cooperative multi-tasking, nor did they have anything resembling a 32-bit filesystem. FAT32, Microsoft's 32-bit file system, didn't come along until Windows 95; prior to that they had FAT16.
Additionally, starting with the 286 you could have more than 640k of RAM. The 286, IIRC, had a 24-bit address space and could therefor address up to 16 megabytes when running in 16-bit protected mode, but even in its protected mode still suffered from the horrid segmentation model that so annoyed programmers writing software for Intel's earlier x86 CPUs. Intel's poor segmentation system didn't become a thing of the past (or at least something you could ignore) until the 386 and its 32-bit protected mode.
Slashdot Mirror
Let's see, there are really only two alternatives:
1)The program allocates some memory, uses it, and then immediately deallocates it. Since the program might not need it for very long (possibly an object-oriented system like you mentioned), this would fall under the "rapidly allocates and deallocates" category.
OR
2)The program allocates some memory, and then waits until it exits (or some other, later time) to deallocate it.
As far as the first line in my post goes, I was trying to imply that any properly written program would be affected at least somewhat by this bug, since only bad programmers will rapidly allocate memory without also (potentially rapidly) deallocating it right after using it.
I think you need to reread my post. Nowhere did I say that programs that allocated a block and then didn't deallocate it until a good while later would be affected. Rather, I said just the opposite.
He was making a joke.
Think about it for a bit; you'll get it eventually.
A program that frequently allocates memory but doesn't deallocate it until much later is either broken, poorly written, or (most likely) both.
I think that this will hit programs that try to be at least somewhat memory efficient the hardest, because those programs will deallocate a block of memory as soon as they're done using it instead of waiting until much later to do it, leaving room for a potential slowdown.
That's how all viruses reproduce. They latch onto a host cell and inject their own DNA into it. The virus DNA then proceeds to hijack the cell's reproductive mechanisms, forcing it to make more viruses inside itself until the cell becomes so full of these new viruses that it bursts open and dies, thereby unleashing the newly made viruses.
No, no, no, if you really want to hijack the name, call it Gnutella2.1 or something.
Basically, because it gives the unwary the impression that it's just an updated version of the same protocol, even though in reality it isn't.
"Doom on you! Doom on you! Doom on you!"
From what I've gathered, Steve Jobs was at the Intel conference representing Pixar, not Apple. As you may remember, Pixar recently began to drop their Sun-powered render farm in favor of a cheaper Intel-powered one. This had nothing to do with Macs vs PCs, Motorola vs Intel, etc.
;) ), and those customers would drop Apple if they thought it was "just another PC."
Similarly, while Apple and Motorola certainly haven't been getting along for a good while now, it doesn't make sense for Apple to switch to a CPU with an entirely new instruction set. Regardless of whether or not OS X runs on x86, all of the Mac OS X software would have to be ported.
What's more, Apple would lose a lot of their customer base, because there's a certain air of eliteness that comes with using a Mac (or, at least there is in the minds of some Mac users
And yet more still, it would be like SGI's Intel boxes: nobody wants to pay through the nose for x86 (c'mon SGI, $30 thousand for an Intel box? No thanks), especially when they can get it for cheaper by ordering it from Dell, HP, or building it themselves.
No, I think Dvorak is just being his usual idiot self. Besides, hasn't Apple already announced (maybe not officially) that they're going with IBM's 970 PPC processor? That would certainly make more sense, since Mac developers wouldn't have to port their software to a whole new architecture, and would only require a new motherboard and some small changes to the OS to handle the 64-bit pointers and such.
The black hole of the story, with billions of solar masses, is probably less dense than water...
Actually, no. From what I understand, the "size" of the black hole is actually the size of the event horizon, since the black hole itself has a size of zero and infinite density.
I'd be willing to bet that the Sci-Fi Channel execs had a hand in that ending.
*SPOILERS BELOW*
By killing off the two main characters, they can say, "Well, see, we can't continue the show, because Crichton and Arynn are dead." And then either a)the producers snuck in the "To Be Continued" so that the fan outrage over not getting to find out if they really are dead would get the show another season/movie/miniseries to resolve it, or that b)the execs at the Sci-Fi Channel were hedging their bets in case the fans were somehow able to resurrect the show.
Just my two cents.
It is my understanding that black holes emit a very small amount of radiation, and do it very slowly. Therefore, the black hole would have to be very small (thereby having very little gravity) for this to take place.
Besides, a black hole would have to swallow all the matter within its reach before it could shrink to the size necessary for the effect you describe to take place, because the more matter it takes in, the bigger it gets. And since it would have already taken in all the matter it can get ahold of, the effect you mention wouldn't happen.
Well, funny posts aside, as scales get more and more accurate, we find that the old ways of measuring how much something is (such as the mass of one kilogram) aren't accurate enough (possibly a small amount of variance that the old instruments couldn't measure, but the newer, more precise scales can).
;)
So, we need new, more precise methods of defining things like how much mass one kilogram has.
Or something like that
The headline talks about "weighing" the black hole, so I used the word weight. But I knew I'd get reamed if I didn't mention that it's actually the mass of the black hole, not its weight, so I put that in there too. So much for hedging my bets.
But you're right. I should've written mass and left it at that.
A Google search reveals that you're right about DOS 4.0.
I was right about DOS 5.0, though.
Technically, FAT32 didn't resemble a 32-bit filesystem either; that came when fdisk 32, which actually enabled the 32-bit goodness at the partition level, was released with Win98 OEM version D.
All fdisk can do is make and delete partitions. The reason you needed the fdisk that came with Windows 95 OSR2 or Windows 98 to create FAT32 partitions was because MS didn't give it the ability to create partitions labelled as using the FAT32 filesystem until then. But that is the end of fdisk's involvement.
FAT32 was, in fact, a 32-bit filesystem. It stored things like disk block numbers, etc., in 32-bits. Microsoft's little secret, though, was that only 22 of those 32 bits were actually used. The other 10 were empty, or something like that.
You did not recall correctly. It's 20-bit. 5 nibbles. And yes, the segmentation was horrible.
Then why does my old 286 have 2 megabytes of RAM? 20 bits of address space means that you can access only 1 megabyte at most, which is what the 8086, 8088, and 186 had. Therefor, the 286 would have HAD to have more that 20 address lines (although, as with modern Intel processors, only 20 of the address lines are available when running in real-mode).
To support this, I kindly refer you to this page.
1)I stand corrected.
2)I'd rather have an OS without the bugs and quirks, thanks. But most programs still ran just fine on DR-DOS, so the point is moot. That, and I was too poor to have very many games, so I didn't really notice.
Win 3 supported every feature of the 386 processor. It could run 32 bit code (although most of the code was 16 bit for compatibility). It could run DOS programs in V86 mode. It supported 4Gb of RAM. That's pretty much every 386 feature accounted for.
You could only run 32-bit programs with the Win32S (yes, there's an S on there) addon. Most programs didn't use it.
4GB of RAM? Really, that's quite astonishing, considering that not even Windows 95 supported that much.
DOS programs in V86 mode? I'll give you that one. Keep in mind that I said Windows 3.1 supported some of the 386's features.
The article claims that DOS tasks where pre-emptively multitasked. This is correct. I thought it was true for 2.0/386 as well, though, but I'm not certain, having never actually used that (I only ever used 2.0 on a 186).
I never stated otherwise. But for all the non-DOS programs, the versions of Windows prior to Windows 95 (and excepting Windows NT) used cooperative multitasking.
No, I am completely on-base.
1)I didn't say otherwise. You, however, (correctly) state later in your own post that Win95 did in fact thunk to 16-bit code in some places.
2)Again, where did I say otherwise?
3)Where do we disagree on this point?
And MS didn't have anything to do with DOS until version 6.2 or whatever.
Not even close. When MS heard that IBM was designing the PC, and looking for an OS for it, they bought a dinky little OS that nobody had ever heard of called QDOS (Quick and Dirty Operating System), which was itself an x86 hack/clone of the popular CP/M operating system, and made some changes to it so that it would run on a PC.
IBM had been hoping to use actual CP/M, but the company that made it, Intergalactic Digital Research (the Intergalactic part was eventually dropped), was too slow off the mark in doing the x86 port (CP/M was originally for the Z80 microprocessor), so they picked Microsoft with their "new" "Disk Operating System". IBM called it PC-DOS, and shipped it themselves for a while, but all development beyond QDOS was done by Microsoft. MS, of course, also licensed DOS to companies making PC clones, such as Compaq, and eventually (at something like DOS 3.0) started selling it to the public themselves.
Intergalactic Digital Research eventually finished the x86 PC port of CP/M (in something like 1982), but by then it was too late. They tried to pull a reverse-Microsoft on Microsoft by marketing their own version of DOS called DR-DOS (they had dropped the Intergalactic part of their name), but Microsoft had released Windows by then, and were able to out-Microsoft Digital Research by making Windows 3.1 only work on MS-DOS and spreading FUD that DR-DOS wasn't as good as MS-DOS (quite the opposite in reality, though; DR-DOS was vastly more stable than MS-DOS, and the only reason that Microsoft even thought about developing MS-DOS beyond version 3.0 was to compete with Digital Research; their marketing department announced MS-DOS 4.0 before any of the design people had even considered it, and later did the same with MS-DOS 5.0).
Does this look familiar?
The article seems to have taken most of its text from that microsoft.com link.
It used preemptive multitasking for MS-DOS apps, because there would be no other way. For everything else, though, it did cooperative multitasking.
I have two things to say about that:
1)From the article: "Many longtime PC users trace Windows to the 1990 release of Windows 3.0, the first widely popular version of Windows and the first version of Windows many PC users ever tried. But Microsoft actually released the first version of Windows six years earlier, in 1985." You'd think that Microsoft of all companies could afford to hire somebody who can do basic arithmetic.
2)The text in the article seems to have been taken almost verbatim from the microsoft.com link you posted. No wonder the entire article seemed like it had been written by a marketroid.
DOS was 16-bit real-mode. Windows 95 was 32-bit protected mode, with some 16-bit DOS underpinnings for backwards compatibility, and so that a complete rewrite wouldn't be necessary.
Of course, if DOS had been a decent operating system this wouldn't have been so bad, but DOS was a port of a knock-off of an OS designed for computers with only 64K of RAM. It should have died the minute the 286 came out, or at least when the 386 came out.
I didn't say that Windows 3.1 didn't support the 386, just that it didn't fully support it (meaning no 32-bit protected mode, no 32-bit apps without the special Win32S addon, etc.), contrary to what the article says.
You have to preemptively multi-task a DOS program running in virtual real-mode if you want to run more than one of them, since DOS programs don't have any cooperative multitasking capabilities (cooperative multitasking is basically a system where the processes have to tell the OS, "Ok, go ahead and let another process have a turn now"; preemptive multitasking is where the OS sets a timer (say, for 20 milliseconds), and when this timer is up a specific interrupt is generated, and this interrupt is the kernel's process scheduler, which then resets the timer and switches to a different process, in other words the OS decides when it's time to switch processes, not the processes themselves). OTOH, regular Windows 3.1 programs were still cooperatively multi-tasked, which is retarded and can be annoying for the users (ask anybody who's ever tried to copy a 100 megabyte file on an OS with cooperative multitasking).
This article seems to have some inaccuracies.
Namely that versions of Windows before Win95 didn't fully support the 386 (dunno 'bout NT, never used it), despite what the article claims, still had worthless (and error-prone) cooperative multi-tasking, nor did they have anything resembling a 32-bit filesystem. FAT32, Microsoft's 32-bit file system, didn't come along until Windows 95; prior to that they had FAT16.
Additionally, starting with the 286 you could have more than 640k of RAM. The 286, IIRC, had a 24-bit address space and could therefor address up to 16 megabytes when running in 16-bit protected mode, but even in its protected mode still suffered from the horrid segmentation model that so annoyed programmers writing software for Intel's earlier x86 CPUs. Intel's poor segmentation system didn't become a thing of the past (or at least something you could ignore) until the 386 and its 32-bit protected mode.