I live in Toronto, and during a recent snowstorm, the signal I was watching went out (on Star Choice).
I checked the dish, it was pretty snowy, so I stuck the broom out the window and gently dusted it off.
Signal was still out. Dumb me then tried another channel. Everything except CBC in Alberta was fine.
I figured the station's uplink had gone dead, which also would have killed the cable feed for that station. (Except it was in Calgary, so you can't get it on cable here. Gotta love multiple timezones of the same network feed for things you didn't know will be on.)
The really funny part is, the signal went out while the show was making fun of the miserable weather we've been having--fake static on the picture and things. It was a taped show (Air Farce).
There was a 20MHz "RSC" (RISC single chip) processor, used in the RS/6000 model 220. Back when the POWER CPU was a multiple-chip board, this was a single-chip implementation of POWER. Circa 1992.
Some quick googling shows that the RSC series is where the RAD6000 came from, "IBM RS/6000, predecessor to the PowerPC,...."
And the specs seem to be 25 MIPS at a 25 MHz clock; which is a pretty bad ratio for an RS/6000.... no wonder the model 220 seemed so horribly slow.
The 601 (featured in the model 250) rocked by comparison; they even gave the expensive multi-chip POWER CPUs at the same clock rate (67 MHz) a good run. The big POWERs would win by having 256 bits to memory instead of a measly 64, but it was close. (The POWER2, though... wow, that was a nice box for its day.)
Windows does, indeed, not have fork(). And I'm going by the Platform SDK manual for NT, 2K, and 9x.
fork() does not just create a child process. It creates a child process running the same program; read-only segments are used as-is, and read-write segments are copied. Open file descriptors are remain open in the child. (Modern systems do copy-on-write for performance, since fork() is often followed by exec(), which loads a new process.)
You can't bolt on fork() from userspace; not efficiently. It needs too much access to the page map tables. Cygwin tries, and it is slow. If you want a program to run well on Windows, you're much, much, much better calling CreateProcess directly, rather than trying to keep fork()/exec() shared with your POSIX implementation.
Instead of using fork(), people write multithreaded applications on Windows. Threads are even faster to initialize, but instead of just sharing the read-only segments, they also share read-write segments (and open file descriptors and all that other stuff). So you have to do your resource management differently. But that's OK, and there are POSIX threads, so everything is still a pain to port anyway.
Of course, threads can communicate via their shared read/write memory, which fork()ed processes cannot do (without using the shm* IPC functions); so they're much better for some classes of problem. But for "launch and run 'till done" tasks, fork() is great, as it gives you process isolation among your workers.
There's a reason why VisualStudio's runtime has a spawn() function.
Someone else provided a link to the list of Apple-supported players; I'd like to chime in that with the Missing Sync package, iTunes "recognizes" my Sony Clie as an MP3 player, and will happily copy files to it.
(And without Missing Sync, you don't get to use a Clie on a Mac. Thanks, Sony.)
There's a plugin framework in iTunes; if Apple doesn't support your player, you [the vendor] can write a little plugin to hook it up.
It is copyright that causes the GPL to 'attach' to the derivative work. If it is not considered a derivative work under copyright, the GPL does not apply. In fact, it cannot apply--it works entirely within copyright; that's the point of the article.
If I put a few pages from your book in my book, the judge can stop me from distributing the whole thing.
If I copy a few lines from your song into my song, the judge can stop me from distributing the whole thing.
And so on. The GPL lets you do things copyright law does not.
It is more restrictive than BSD, but it is still more generous than simple copyright.
No, you do not have permission to release the GPLed work under your proprietary license. The GPL only allows you to release a derivative work under the GPL.
You would have to get permission from those copyright holders as well, which you may very well get by offering them something like a portion of your revenue... which would then plunk you down into contract territory for the proprietary version.
It's all basic copyright: the only thing giving you permission to use the work you created a derivative of is the GPL. You don't have to use that code, you can always implement your own. Or, if the origins are well-enough known, negotiate a different license with the copyright holder. (This should be possible with most FSF-owned works, as they require copyright assignment of patches. It is probably impossible with the Linux kernel....)
However, despite that, I am completely offended by the presumption-of-guilt that this levy relies on, especially given the amount of levied media I use for something other than music.
Since I learned about these levies and the linked-to section of the Copyright Act, I've been more than willing to copy anything, to make up for all the levies I've paid without doing unauthorized copying.
Noting that there is a difference between "unauthorized" and "illegal" copying....
- It consumes an extra process slot by doing command followed by exit (rather than exec command). This also prevents a program from finding abnormal termination status from the command (i.e., signalled instead of exited). - It hides any error messages from the user. - It does not handle an argument with spaces or other Weird Characters in them.
Pretty bad, 3 errors for 2 lines of code. No wonder they want to keep it confidential.
Re:How can IBM provide what SCO is requesting?
on
SCOrched Earth
·
· Score: 1
Last time I looked (1999) IBM was still running their in-house product, CMVC/6000, to manage the AIX source code. Checking the revision strings on the AIX 5.2 system headers, they still are. (Despite CMVC having been discontinued for about 5 years now.)
You want the source for a rev. of AIX, they've got it. I would be extremely surprised if there was any trouble fetching AIX v1 (for PS/2) or v2 (for RT/PC).
Keeping source code history is so small an amount of data, there's no reason not to. Where I work now, the daily builds and packaging images sit on a terabyte-and-a-half array. The entire CVS server fits on a single 18 GB drive, and it's only half full--and that's at least 7 years of history.
Of course, all that matters if all you need is bandwidth. If latency matters, and I'm talking about file operations, Sun's going to kill you. Even Solaris on Intel is a better choice than Solaris on SPARC, but Linux on Intel leaves them both in the dust. (I was only concerned about latency, as shown by CVS I/O, so that's all I tested; maybe the bandwidth wasn't there on the SPARC either--that would really suck for Sun.)
And yes, I'm comparing expensive fibre arrays on the Solaris SPARC to Ultra320 and UDMA133 on the Linux Intel machine.
Make sure you spend your money where it matters, big iron isn't for every job, sometimes your better with a lot of cheap little boxes. Ask Google.
While this isn't drive imaging software per se, it does handle the restore a system to an operating state feature. Dantz Retrospect has an add-on which they say can make a bootable disaster-recovery CD which will regenerate any client of the Retrospect backup server. AFTER the client has failed.
I've never used that feature, as my primary use of Retrospect is on Macintosh; I have a Windows client, but have not had to try to regenerate it following a total system loss. (And I don't have the add-on to do it anyway.)
But the rest of Retrospect is common across Windows and Mac. (Disaster recovery on Mac seems to be a lot easier.) The important part for this discussion, in addition to the 'disaster recovery CD' add-on, is the way it does a so-called 'snapshot' when it takes an incremental backup. This lets you get both the speed of doing incremental backups, plus the ability to restore a system to precisely the contents it had at that time. (So basically, it can handle deletes too, so it doesn't need to restore files from the full backup which were deleted when a later backup took place.)
I bought Retrospect for Macintosh after Norton 'Crash Gaurd Causes Crashes' Utilities removed their backup/restore software in a newer version. (Fortunately, the id10ts at Symantec offered a satisfaction guarantee on software. The store didn't believe it, they had to place a toll call to Symantec to find out they weren't kidding.)
I've been using Retrospect for Macintosh since System 8 came out, through OS X, and now with a Windows 98 client. It's my very favourite backup program, and what's more, the restores work.
Just need to get the upgrade to 5.1 for Mac so I don't need a separate backup system for my Linux box.
By the way, it sounds like you've got an axe to grind about FireWire though.
Did you USE any of the early FireWire to IDE bridges?
You'd have an axe to grind too; the OP was very clear about the pre-Oxford911 chipset bridges. They sucked. They sucked camels through soda straws. They were slow; they would crash the whole bus; they would not work with each other for no apparent reason. Total crap.
Keyspan's USA-19QW works a treat on OS X, Classic (within OS X) and Win98SE.
The only annoying thing is the "You unplugged the device while an application was using it" pop-up if you're using it with Palm Hotsync and you unplug the dongle. Normal applications that actually close the device when they're done won't have this problem.
I guess the finance people figure that the tech people don't know anything since they totaly blew the.com thing.
Maybe I'm misreading that, but I thought the.com thing was a finance-people thing; all the tech people I know of just stood back and shook their heads at it all.
Well, the few who got in on some juicy IPOs of valueless companies giggled all the way to the stock broker, but after that they just laughed at everyone else.
Without an intimate knowledge of the drive geometry, how can reordering reads/writes be done efficiently?
The RAID controller would re-order I/O by grouping "near" sector addressess together. (SCSI has always used LBA-type addressing, IDE can't live without it now.) The on-drive controller has to be set up so that is a reasonable thing to do, that writing block 1, 2, 3, 4, 5 minimizes head seek and rotational latency. Sometimes you're going to have a stripe cross a cylinder boundary; but that's just one crossing, you won't be dancing back and forth over it while writing that portion of the stripe.
TCQ allows the drive controller circuitry itself (downstream of a SCSI controller) to re-order writes; the host adapter issues writes for blocks 1, 1001, 2, and 1002, and the drive controller says, "done 1, 2, 1002, 1001." The controller circuitry does know the geometry; it has to.
So sure there are issues with not knowing the geometry, but it turns out to not really matter after all; with the disk buffer to absorb the track-to-track delay, you're fine. (As long as seeks don't dominate your I/O, which is what the test was about.)
I think the responsibility lies MOSTLY with the application developers who want you to buy a new copy of their product whenever Apple releases a major update.
I have never purchased another Symantec product ever since Norton Utilities displayed this message:
This version of Norton Utilities is not supported on System 8.
And that was for the backup/restore program, you know, which reads and writes files? Like THAT is something that changes release to release?
(The winner: The new version didn't even have the backup/restore program, so there was no point in buying it. Dantz got a new customer that day.)
Most modern processors, 68000 era and later, have a 'HALT' instruction which stops most of the internal 'ticking' of the CPU until an interrupt is received. On a CMOS CPU, your power use can go to approximately zero.
Check the boot messages on Linux; see the one where it says "Checking 'hlt' instruction"? That's what that is. Without hlt, the kernel has to do a no-op loop when there's nothing to run.
I believe all Windows NT versions (3.0 through 5.1 oops I mean XP) use hlt; there was some fuss about the DOS-based Windows not using it, but I don't care enough to look it up.
Slashdot Mirror
Except in the case of a mixture of text and numbers. A lot of people don't use tab at all, let alone properly, and just space over.
I live in Toronto, and during a recent snowstorm, the signal I was watching went out (on Star Choice).
I checked the dish, it was pretty snowy, so I stuck the broom out the window and gently dusted it off.
Signal was still out. Dumb me then tried another channel. Everything except CBC in Alberta was fine.
I figured the station's uplink had gone dead, which also would have killed the cable feed for that station. (Except it was in Calgary, so you can't get it on cable here. Gotta love multiple timezones of the same network feed for things you didn't know will be on.)
The really funny part is, the signal went out while the show was making fun of the miserable weather we've been having--fake static on the picture and things. It was a taped show (Air Farce).
There was a 20MHz "RSC" (RISC single chip) processor, used in the RS/6000 model 220. Back when the POWER CPU was a multiple-chip board, this was a single-chip implementation of POWER. Circa 1992.
...."
Some quick googling shows that the RSC series is where the RAD6000 came from, "IBM RS/6000, predecessor to the PowerPC,
And the specs seem to be 25 MIPS at a 25 MHz clock; which is a pretty bad ratio for an RS/6000.... no wonder the model 220 seemed so horribly slow.
The 601 (featured in the model 250) rocked by comparison; they even gave the expensive multi-chip POWER CPUs at the same clock rate (67 MHz) a good run. The big POWERs would win by having 256 bits to memory instead of a measly 64, but it was close. (The POWER2, though... wow, that was a nice box for its day.)
Windows does, indeed, not have fork(). And I'm going by the Platform SDK manual for NT, 2K, and 9x.
fork() does not just create a child process. It creates a child process running the same program; read-only segments are used as-is, and read-write segments are copied. Open file descriptors are remain open in the child. (Modern systems do copy-on-write for performance, since fork() is often followed by exec(), which loads a new process.)
You can't bolt on fork() from userspace; not efficiently. It needs too much access to the page map tables. Cygwin tries, and it is slow. If you want a program to run well on Windows, you're much, much, much better calling CreateProcess directly, rather than trying to keep fork()/exec() shared with your POSIX implementation.
Instead of using fork(), people write multithreaded applications on Windows. Threads are even faster to initialize, but instead of just sharing the read-only segments, they also share read-write segments (and open file descriptors and all that other stuff). So you have to do your resource management differently. But that's OK, and there are POSIX threads, so everything is still a pain to port anyway.
Of course, threads can communicate via their shared read/write memory, which fork()ed processes cannot do (without using the shm* IPC functions); so they're much better for some classes of problem. But for "launch and run 'till done" tasks, fork() is great, as it gives you process isolation among your workers.
There's a reason why VisualStudio's runtime has a spawn() function.
Someone else provided a link to the list of Apple-supported players; I'd like to chime in that with the Missing Sync package, iTunes "recognizes" my Sony Clie as an MP3 player, and will happily copy files to it.
(And without Missing Sync, you don't get to use a Clie on a Mac. Thanks, Sony.)
There's a plugin framework in iTunes; if Apple doesn't support your player, you [the vendor] can write a little plugin to hook it up.
By this logic, copyright is viral.
It is copyright that causes the GPL to 'attach' to the derivative work. If it is not considered a derivative work under copyright, the GPL does not apply. In fact, it cannot apply--it works entirely within copyright; that's the point of the article.
If I put a few pages from your book in my book, the judge can stop me from distributing the whole thing.
If I copy a few lines from your song into my song, the judge can stop me from distributing the whole thing.
And so on. The GPL lets you do things copyright law does not.
It is more restrictive than BSD, but it is still more generous than simple copyright.
No, you do not have permission to release the GPLed work under your proprietary license. The GPL only allows you to release a derivative work under the GPL.
You would have to get permission from those copyright holders as well, which you may very well get by offering them something like a portion of your revenue... which would then plunk you down into contract territory for the proprietary version.
It's all basic copyright: the only thing giving you permission to use the work you created a derivative of is the GPL. You don't have to use that code, you can always implement your own. Or, if the origins are well-enough known, negotiate a different license with the copyright holder. (This should be possible with most FSF-owned works, as they require copyright assignment of patches. It is probably impossible with the Linux kernel....)
They already don't have it both ways.
However, despite that, I am completely offended by the presumption-of-guilt that this levy relies on, especially given the amount of levied media I use for something other than music.
Since I learned about these levies and the linked-to section of the Copyright Act, I've been more than willing to copy anything, to make up for all the levies I've paid without doing unauthorized copying.
Noting that there is a difference between "unauthorized" and "illegal" copying....
Also note:
- It consumes an extra process slot by doing command followed by exit (rather than exec command). This also prevents a program from finding abnormal termination status from the command (i.e., signalled instead of exited).
- It hides any error messages from the user.
- It does not handle an argument with spaces or other Weird Characters in them.
Pretty bad, 3 errors for 2 lines of code. No wonder they want to keep it confidential.
Last time I looked (1999) IBM was still running their in-house product, CMVC/6000, to manage the AIX source code. Checking the revision strings on the AIX 5.2 system headers, they still are. (Despite CMVC having been discontinued for about 5 years now.)
You want the source for a rev. of AIX, they've got it. I would be extremely surprised if there was any trouble fetching AIX v1 (for PS/2) or v2 (for RT/PC).
Keeping source code history is so small an amount of data, there's no reason not to. Where I work now, the daily builds and packaging images sit on a terabyte-and-a-half array. The entire CVS server fits on a single 18 GB drive, and it's only half full--and that's at least 7 years of history.
These would be the same sort of people who paid for all those .com stocks, right?
They'll get what's coming to them.
Of course, all that matters if all you need is bandwidth. If latency matters, and I'm talking about file operations, Sun's going to kill you. Even Solaris on Intel is a better choice than Solaris on SPARC, but Linux on Intel leaves them both in the dust. (I was only concerned about latency, as shown by CVS I/O, so that's all I tested; maybe the bandwidth wasn't there on the SPARC either--that would really suck for Sun.)
And yes, I'm comparing expensive fibre arrays on the Solaris SPARC to Ultra320 and UDMA133 on the Linux Intel machine.
Make sure you spend your money where it matters, big iron isn't for every job, sometimes your better with a lot of cheap little boxes. Ask Google.
I've never used that feature, as my primary use of Retrospect is on Macintosh; I have a Windows client, but have not had to try to regenerate it following a total system loss. (And I don't have the add-on to do it anyway.)
But the rest of Retrospect is common across Windows and Mac. (Disaster recovery on Mac seems to be a lot easier.) The important part for this discussion, in addition to the 'disaster recovery CD' add-on, is the way it does a so-called 'snapshot' when it takes an incremental backup. This lets you get both the speed of doing incremental backups, plus the ability to restore a system to precisely the contents it had at that time. (So basically, it can handle deletes too, so it doesn't need to restore files from the full backup which were deleted when a later backup took place.)
I bought Retrospect for Macintosh after Norton 'Crash Gaurd Causes Crashes' Utilities removed their backup/restore software in a newer version. (Fortunately, the id10ts at Symantec offered a satisfaction guarantee on software. The store didn't believe it, they had to place a toll call to Symantec to find out they weren't kidding.)
I've been using Retrospect for Macintosh since System 8 came out, through OS X, and now with a Windows 98 client. It's my very favourite backup program, and what's more, the restores work.
Just need to get the upgrade to 5.1 for Mac so I don't need a separate backup system for my Linux box.
The Toronto mayoral race has 44 candidates, so you don't need to go that far afield; just the next town over really.
"...and if you think they're all a bunch of backstabbing, lying, cheating scumbags, press 45#."
You've never used a FireWire CD burner on Linux, have you? It'll panic even if you only put plain white toast in the drive.
Did you USE any of the early FireWire to IDE bridges?
You'd have an axe to grind too; the OP was very clear about the pre-Oxford911 chipset bridges. They sucked. They sucked camels through soda straws. They were slow; they would crash the whole bus; they would not work with each other for no apparent reason. Total crap.
Keyspan's USA-19QW works a treat on OS X, Classic (within OS X) and Win98SE.
The only annoying thing is the "You unplugged the device while an application was using it" pop-up if you're using it with Palm Hotsync and you unplug the dongle. Normal applications that actually close the device when they're done won't have this problem.
Maybe I'm misreading that, but I thought the .com thing was a finance-people thing; all the tech people I know of just stood back and shook their heads at it all.
Well, the few who got in on some juicy IPOs of valueless companies giggled all the way to the stock broker, but after that they just laughed at everyone else.
And, as another poster said, there is a levy on data CDRs, just a lower one than on music CDrs.
But there's nothing that requires you to only copy music to levy-paid media; you can copy it to anything you like. They're separate parts of the law.
The RAID controller would re-order I/O by grouping "near" sector addressess together. (SCSI has always used LBA-type addressing, IDE can't live without it now.) The on-drive controller has to be set up so that is a reasonable thing to do, that writing block 1, 2, 3, 4, 5 minimizes head seek and rotational latency. Sometimes you're going to have a stripe cross a cylinder boundary; but that's just one crossing, you won't be dancing back and forth over it while writing that portion of the stripe.
TCQ allows the drive controller circuitry itself (downstream of a SCSI controller) to re-order writes; the host adapter issues writes for blocks 1, 1001, 2, and 1002, and the drive controller says, "done 1, 2, 1002, 1001." The controller circuitry does know the geometry; it has to.
So sure there are issues with not knowing the geometry, but it turns out to not really matter after all; with the disk buffer to absorb the track-to-track delay, you're fine. (As long as seeks don't dominate your I/O, which is what the test was about.)
To that 1 in 23, you can get Solitaire.
Because for some operations, such as a CVS server, Solaris on x86 is rather faster than Solaris on SPARC. Linux is faster still.
I have never purchased another Symantec product ever since Norton Utilities displayed this message:
And that was for the backup/restore program, you know, which reads and writes files? Like THAT is something that changes release to release?
(The winner: The new version didn't even have the backup/restore program, so there was no point in buying it. Dantz got a new customer that day.)
Most modern processors, 68000 era and later, have a 'HALT' instruction which stops most of the internal 'ticking' of the CPU until an interrupt is received. On a CMOS CPU, your power use can go to approximately zero.
Check the boot messages on Linux; see the one where it says "Checking 'hlt' instruction"? That's what that is. Without hlt, the kernel has to do a no-op loop when there's nothing to run.
I believe all Windows NT versions (3.0 through 5.1 oops I mean XP) use hlt; there was some fuss about the DOS-based Windows not using it, but I don't care enough to look it up.
Here's a better link to the private copying provision of the Canadian copyright law.