Check out this, and this. The avifile-player package in Debian is in the main section (thus whoever put it there believes that it meets the Debian Free software guidelines, which are very stringent in guaranteeing freedom of the source code and everything). I can use aviplay to watch DivX vids I've downloaded, without any windows DLLs or any other binary only or otherwise non-free software. (I can download them with Free gnutella software too: napshare:)
> Everyone claims that this was sophisticated, planned for months, etc.
That's exactly what I thought when I heard the dudes on the news going on about how this attack required massive amounts of resources, etc. I have no idea how they justify that. They talk about it as if that fact were self-evident. Just because it killed a lot of people doesn't mean it required a lot of resources.
If I was trying to pull this off, I'd get four groups of hijackers. We'd try to get some stuff we could use as weapons onto the planes. We'd call each other on cell phones to make sure all four planes were ready for hijacking. If not, try again the next day. If we were ready, just take over and fly the planes into something important. How is this hard? With an internet connection, a laptop, and good encryption, the authorities wouldn't have a clue.
I'm not actually a terrorist, so maybe I'm missing something here... The only thing I can think of is that maybe you would need a big organization to get enough suicide hijackers, and ones who could fly as well. With parachutes, some of the hijackers might even figure out a way to pull it off and survive.
The only way to defend against this kind of thing once the plane is hijacked is to have SAM sites protecting important buildings. I don't think you could scramble fighters fast enough, unless you have a huge no-fly zone over your important buildings.
Even more panic and havoc could have been caused by spreading some nasty infectious diseases. Viruses keep killing indefinitely, and with an incubation period. You don't even know if you've got it until you break out. This would cause _real_ panic. The West Nile virus caused some panic, and it's barely a threat at all. Imagine Ebola outbreaks on a couple stock traders who'd been in the stock exchange that day...
This terrorist attack hit symbols of power, and killed a couple thousand people. I just hope that nobody tossed some Anthrax or something into the mix. We won't know for a while whether any bio-weapons were used, because of the incubation period.
Terrorist threats are the major threat to the world these days, AFAIK. None of them would be insane enough to use ICBMs, because they couldn't deny involvement very easily if the missile was launched from their base. Screw missile defence. (It's not useless, since a missile could be fired by accident or something. I just don't think it's worth the money it would cost, and that that money could be better spent on health care and stuff.)
What about the White House? Wouldn't that be a likely target? If not, what does it say about the terrorists' goals that they _didn't_ target the White House?
The targets they did hit are symbols of American power, economic and military. Do they think that American political power is not as important to the public as money and guns? Are they right in that thought?
> it was under the GPL, and the author decided to release a major new version under a shareware license.
Yes, that's exactly what I was refering to.
Thanks for the info. I haven't played any Xevil for a year or more, so I wasn't aware of how things had gone. I remember some people had forked the last GPL release, etc. I'm glad to see Steve Hardt is back to the GPL version now.
Hmm... apt-get install xevil... re-enters a world of chainsaws, caffeine, the Altar of Sin:), and becomes Satan's Earwax remover... I love the smell of hot napalm in the morning.
(disclaimer: I haven't proof read this to make sure I said stuff in the right order without repeating myself. My ideas are in here somewhere:)
> And you couldn't possibly stand to allow them to think for themselves and form their own opinions, could you now....
Don't get too upset, Brett. I hadn't heard of you before this story, and I found it useful to know what your views on Free licenses in general were. If the OAL was criticized by RMS (whose views I'm familiar with), I would suspect that the OAL had failed to achieve what it intended. Given what Bruce says about your views (anti-GPL in general), I am confident in taking your criticism of the OAL as a criticism aims of the OAL. (For the most part: You have some good points about how the OAL doesn't allow for GPLing the song, but selling recordings by specific artists, or whatever. However, a minute's thought will lead one to the realization that you are not stuck between using the OAL or going with the normal proprietary music distribution model. If you want to do something that this version of the OAL doesn't work for, then take the useful clauses from the OAL, and change the problematic one/ones to say what you want. It's not hard to change the license to say that you're allowed to distribute this recording of the song (but not re-record it without an ok), or that it can be sampled without viral effect, or whatever.)
Bruce's commentary on your views didn't hypnotize me into submission, sheep-like. Don't worry. I'm still thinking for myself, just like most (err, some...) other/.ers. I appreciate having more information.
>> Yes, I think that taking public software private is a selfish act.
> Again, Bruce, the same misleading propaganda.
How is that propaganda? That's his own personal opinion. Some public->private transitions have been pretty nasty. In the case of OpenDivX, the company deceived the contributors and testers into thinking that OpenDivX would be the main effort. In reality, they were working on their own proprietary DivX codec, incorporating their own ideas as well as all the ideas contributed by OpenDivX contributors. Even without dirty underhanded moves like that, it still seems selfish to Bruce and myself. Another case I remember is the game Xevil. It used to be Free software, now it's shareware.
It's hard to argue that taking software private is not selfish, even if it is software that you did most of the work on. Why else would one do it, other than their own self-interest. I suppose the software may improve with people getting more money to work on it, but that's almost a moot point if the improvements are proprietary. As a user of Free software, I am used to the people developing it being around to receive bug reports, make small fixes when needed, etc. if nothing more. Taking software private makes it abandonware, unless it is popular enough to attract some people willing to keep the project going, as in the case of OpenSSH (which has been very successful, due to the usefulness of the SSH to people who are good at coding.)
Anyway, for better or for worse, taking Free software private is selfish. Everyone has a certain degree of greed. I'm not saying (and I don't think Bruce is saying) that nobody should ever take software private, but I wish people wouldn't do it. They have every right to do so, but I see it as a cop-out. I know you have no problem with proprietary software, etc., I just don't like the way you seem unwilling to be realistic about your views. What I think you are saying is that it's ok for people to be selfish. Fine. Don't try to call it propaganda-busting.
> The GPL claims to be about freedom, but is actually intended to destroy programmers' livelihoods and companies.
No, it's intended to destroy the current business model of software companies. gnu.org says as much in their philosophy pages. As I see it, writing proprietary software is a bad business to get into. This view works for me because I just graduated, and haven't really committed to a career. (I want to design network hardware, or something like that. I do physics as well as CS. If I just did programming, I would want to have people pay me to write software that they needed to accomplish something, not so they could sell copies of it. That may be hard to achieve, and this whole parenthetical bit is way off topic, so I'll drop it.) Proprietary software will take long enough to go away (if it ever does) that current not a lot of career-programmers will be on the street because of the GPL. (The current economic situation has more to do with recent problems than the GPL itself.)
> The GPL does not. The "giving" forced by the GPL is not "giving" at all, because it is compulsory. It does not leave one free to give or not give.
GPLed software is a gift to like-minded programmers, and it is nothing to creators of proprietary software. It is not a gift to everyone, but it is still a gift. If you don't give gifts, you don't get gifts. Makes sense to me.
> And it has done quite well as this. Andy Hertzfeld's company is already gone...
If you're trying to say that you can't build a business on selling support for GPLed software, how do you explain the fact that Cygnus consistently makes a profit. (Probably not a consistent profit, I haven't heard about their profitability the past year either, so correct me if I'm wrong.)
If you're going to reply to me, I should give you a brief outline of my views: I think proprietary software has its uses, but only in "leaf" applications. For example, an Atomic Force Microscope interface program can be proprietary without much harm. (In the case I encountered in the university physics lab a couple summers ago, a Free control program would have been more useful, because we could have make a better tool out of it with more control of the software.) In any case, the AFM control program wasn't a key part of any system other than the one made up of that software and the AFM hardware. OTOH, software like a windowing toolkit (e.g. KDE) really really should be Free. Lots of other software is built upon it, and it wouldn't be possible to use software that required proprietary components on a Free system. (This only matters to ideological purists like myself in the case where the toolkit is free enough to not require royalty payments for distributing software using it, and to allow porting to other computer architectures.) Even KDE wasn't Free enough at first, thus the GNOME project. Basically, general purpose, commonly used software (i.e. the building blocks of a computing environment) should be Free, IMHO. This is important, because it allows anyone to contribute to the evolution of the system.
It was obvious from the tone of the comment he made that he wasn't picking on Hawk, just pointing out that he hadn't dotted all his 'i's, as it were. He was pointing out that doing so is not necessary. Note that he didn't tell Hawk to go away and not come back until he knew HTML or anything. I thought he made his point quite tactfully.
BTW, I agree with his main point, especially his observation that too many lawyers generate more lawyers. The system enters "legal runaway" (once the positive feedback coefficient is greater than one).
I don't write code for Berlin, and I haven't even gotten around to trying it out... Maybe now that it's useful enough for someone to make a.deb of it, I'll give it a shot. Thus, I don't know the answer to your question about invokeNow(), etc.
You could make a progress bar async by having a display thread that drew the bar at a length corresponding to the current amount of progress. (so updates the the bar length that happened while the display thread was blocked would all be skipped, except the last one.)
So, yeah, the drawing thread still blocks, but at least it won't be slowing down your chess analysis on your dual proc machine:) (or on the remote machine, if the server is not on the same machine as the client).
SHM is only used for put/get image stuff, AFAIK.
Local communication happens over Unix sockets (/tmp/.X11-unix/X0, etc.) when DISPLAY=:0 (or:1...). If you set DISPLAY=localhost:0, the xlib will use TCP. (This is occasionally useful when hacking around if you want to run an X client from inside a chroot, where the Unix-domain sockets in/tmp aren't available).
There is a lot of intelligence on the server, so very few actions need to ask the client for help, and those that do are very broad and generic, and so likely wouldn't use up as much bandwidth (as if that's a problem anyway).
Latency is a problem here, as well as bandwidth. On a large-latency link (e.g. CA*Net3, a multi-Gb/s link between several Canadian Universities), the round trip latency for light-in-fiber alone is ~80ms. This means that the program blocks for that long whenever it asks the server to do anything. Displaying a progress bar could slow down progress significantly, unless the process was multithreaded, with a thread dedicated to communicating with the server. (This would make communication asynchrounous.)
OpenSSH has been (and is!) a lot more successful than I think an openGFS project would be. There is a critical need for secure remote access by far more people than there is for good distributed filesystems. There aren't as many Beowulf hackers as there are internet-using hackers. (that's a lot of coders...)
I don't think OpenGFS would reach a critical mass of developers. Of course, the most important consideration is how good the last Free version of GFS was, and how much work it is to keep up with what Sinistra does with their GFS.
I don't want to discourage anyone from working on this; I would love to be proved wrong:) I'm just trying to be realistic.
A short time after I found out about Unix from a CS course at university, I got a PC to use as a Linux workstation, instead of messing with MiNT on my puny Atari (mega 4 STe, ~80MB HD). I knew of MiNT, but I never used it.
Anyway, my major reason for not mentioning MiNT is that AFAIK no Atari ever shipped with MiNT installed. You had to do that yourself. My (possibly incorrect) impression of it was that, like Linux, it was best suited to the geek crowd.
I use ulimit -S -v 400000 to limit the virtual memory size of any one process to ~400MB on my machine with 512MB of RAM, and a bit of swap. This way, if a program decides to eat memory like crazy, it will be stopped before it can deprive any other processes of memory. This way, netscape dies, instead of init.:)
I'm writing a program to do physics simulations (soap froth systems, BTW), and if I goof up and give it command line options that would make it thrash, it effectively runs out of memory _before_ it can make the system thrash.
Since I only set a soft limit, I can jack it up without having to be root or anything. Thus, this is a good thing to put in/etc/profile, where everyone gets it. It won't cause trouble for anyone, because those who need to can override it.
No sane user installs a new kernel without leaving the option of booting the old one. So the user may be pissed off because the effort of compiling and rebooting was wasted, but the system will still be bootable without resorting to that dust-encrusted floppy drive.
(if not, then that's a lesson learned... a simple boot off the RH CD will get you back in business.)
GNU parted is good. It can't move and resize at the same time, so there are some things you can't do with it, but that's in the works.
For shuffling data around, partimage is good. It is basically dump(8), but it will do ext2, FAT, and maybe some others.
I heard about a NIC that plugged into the ACSI port. (for non-Atarians, this is SCSI with s/small/Atari/. It's a similar protocol, and it supports DMA. It's the fastest read-write IO port on the ST. (the cartridge port is on the CPU data & address busses, but the non-programmable MMU bus-errors if you try to write to it. The only write path is through the address bus, reading addresses where the low bits are the data.))
What do you expect when there is a system call to run a function in supervisor mode? (m68k jargon' this is the same as "kernel mode" on processors that were designed to run Real OSes... (m68000 (found in all ST machines) didn't have a programmable MMU. 68020 supported an external one. 68030 and up had it built in.))
The GUI never got around to supporting more than one app open at the same time. (You could have some desk accessories running, though.) Basically, any game that needed a decent frame rate took over the whole computer anyway. Most commercial games came on floppies that you booted from. The base OS (in ROM) just booted the code from the floppy, so the GUI (also in ROM) never had any of its code executed. So not only can any game bring the whole thing down, most commercial games require you to bring your system down before playing them!
Sometimes trying to figure out how to tell it to do something is like playing a Sierra game, though. (give plasma to ugly dude -> you can't do that... That word is not in the Andromedan dictionary... etc.)
Simple stuff is pretty easy to tell MacOS about, but sometimes it's hard to figure out where the controls for something is. (This wouldn't be a problem after plenty of experience, though. Obviously the most important consideration is how easy it is to use for people who use it a lot, so this isn't too bad.)
Nice idea, but Intel already has conditional move, and maybe some other conditional insns. For web serving and compiling, and other integer tasks, a bunch of ARM processors might well do a good job. I wouldn't guess that it would be particularly easy to implement a very fast ARM machine, since the insn set is not simpler than the Alpha's for example, I don't think.
Anyway, it might be worth doing this, but probably not, unless it can be done cheaply. #define X(x,y) x##y
fast, power efficient, x86 compat: choose 2. x86 doesn't pipeline easily. It's complicated to implement in a way that runs fast, as evidenced by everyone who's tried, such as Intel, AMD, or Transmeta. If you wanted power-efficient x86, you'd have to give up on pipelining and stuff, i.e. your chip would be slow, like a 386. #define X(x,y) x##y
If you used Compaq/Digital's optimizing C compiler, you obviously found that the compile took a long time, because the compiler spends a long time scheduling the instructions for the Alpha's pipeline and exact insn execution capability. I read somewhere that the compiler actually simulates an Alpha running the code to see what it can do to make faster code. No wonder it takes a long time, and no wonder the code it generates is so good.
Compilation speed is nice, esp. when developing software, but you can usually get that by turning optimization off. When you're compiling something that will eventually use more CPU cycles than it took to compile, it's ok if the compile takes a long time, as long as the compiler does something useful with that time! #define X(x,y) x##y
Slashdot Mirror
Since you're on FBSD, you'll want to download the source from wherever it came from, or just grab the original tarball from the Debian archive.
Use SVG instead of images. It's like the LaTeX picture environment for HTML.
better than stripping it out: replace it with mouse movements tracing out a peace sign. (or maybe a middle finger...)
Check out this, and this. The avifile-player package in Debian is in the main section (thus whoever put it there believes that it meets the Debian Free software guidelines, which are very stringent in guaranteeing freedom of the source code and everything). I can use aviplay to watch DivX vids I've downloaded, without any windows DLLs or any other binary only or otherwise non-free software. (I can download them with Free gnutella software too: napshare :)
> ... the only two OS options you have for a Powerbook are MacOS and Linux.
You obviously meant to include the qualifier "mainstream". Even at that, you left out NetBSD. (I don't know if BeOS will run on a PB.)
> Everyone claims that this was sophisticated, planned for months, etc.
That's exactly what I thought when I heard the dudes on the news going on about how this attack required massive amounts of resources, etc. I have no idea how they justify that. They talk about it as if that fact were self-evident. Just because it killed a lot of people doesn't mean it required a lot of resources.
If I was trying to pull this off, I'd get four groups of hijackers. We'd try to get some stuff we could use as weapons onto the planes. We'd call each other on cell phones to make sure all four planes were ready for hijacking. If not, try again the next day. If we were ready, just take over and fly the planes into something important. How is this hard? With an internet connection, a laptop, and good encryption, the authorities wouldn't have a clue.
I'm not actually a terrorist, so maybe I'm missing something here... The only thing I can think of is that maybe you would need a big organization to get enough suicide hijackers, and ones who could fly as well. With parachutes, some of the hijackers might even figure out a way to pull it off and survive.
The only way to defend against this kind of thing once the plane is hijacked is to have SAM sites protecting important buildings. I don't think you could scramble fighters fast enough, unless you have a huge no-fly zone over your important buildings.
Even more panic and havoc could have been caused by spreading some nasty infectious diseases. Viruses keep killing indefinitely, and with an incubation period. You don't even know if you've got it until you break out. This would cause _real_ panic. The West Nile virus caused some panic, and it's barely a threat at all. Imagine Ebola outbreaks on a couple stock traders who'd been in the stock exchange that day...
This terrorist attack hit symbols of power, and killed a couple thousand people. I just hope that nobody tossed some Anthrax or something into the mix. We won't know for a while whether any bio-weapons were used, because of the incubation period.
Terrorist threats are the major threat to the world these days, AFAIK. None of them would be insane enough to use ICBMs, because they couldn't deny involvement very easily if the missile was launched from their base. Screw missile defence. (It's not useless, since a missile could be fired by accident or something. I just don't think it's worth the money it would cost, and that that money could be better spent on health care and stuff.)
What about the White House? Wouldn't that be a likely target? If not, what does it say about the terrorists' goals that they _didn't_ target the White House?
The targets they did hit are symbols of American power, economic and military. Do they think that American political power is not as important to the public as money and guns? Are they right in that thought?
> it was under the GPL, and the author decided to release a major new version under a shareware license.
... re-enters a world of chainsaws, caffeine, the Altar of Sin :), and becomes Satan's Earwax remover... I love the smell of hot napalm in the morning.
Yes, that's exactly what I was refering to.
Thanks for the info. I haven't played any Xevil for a year or more, so I wasn't aware of how things had gone. I remember some people had forked the last GPL release, etc. I'm glad to see Steve Hardt is back to the GPL version now.
Hmm... apt-get install xevil
(disclaimer: I haven't proof read this to make sure I said stuff in the right order without repeating myself. My ideas are in here somewhere :)
/.ers. I appreciate having more information.
...
> And you couldn't possibly stand to allow them to think for themselves and form their own opinions, could you now....
Don't get too upset, Brett. I hadn't heard of you before this story, and I found it useful to know what your views on Free licenses in general were. If the OAL was criticized by RMS (whose views I'm familiar with), I would suspect that the OAL had failed to achieve what it intended. Given what Bruce says about your views (anti-GPL in general), I am confident in taking your criticism of the OAL as a criticism aims of the OAL. (For the most part: You have some good points about how the OAL doesn't allow for GPLing the song, but selling recordings by specific artists, or whatever. However, a minute's thought will lead one to the realization that you are not stuck between using the OAL or going with the normal proprietary music distribution model. If you want to do something that this version of the OAL doesn't work for, then take the useful clauses from the OAL, and change the problematic one/ones to say what you want. It's not hard to change the license to say that you're allowed to distribute this recording of the song (but not re-record it without an ok), or that it can be sampled without viral effect, or whatever.)
Bruce's commentary on your views didn't hypnotize me into submission, sheep-like. Don't worry. I'm still thinking for myself, just like most (err, some...) other
>> Yes, I think that taking public software private is a selfish act.
> Again, Bruce, the same misleading propaganda.
How is that propaganda? That's his own personal opinion. Some public->private transitions have been pretty nasty. In the case of OpenDivX, the company deceived the contributors and testers into thinking that OpenDivX would be the main effort. In reality, they were working on their own proprietary DivX codec, incorporating their own ideas as well as all the ideas contributed by OpenDivX contributors. Even without dirty underhanded moves like that, it still seems selfish to Bruce and myself. Another case I remember is the game Xevil. It used to be Free software, now it's shareware.
It's hard to argue that taking software private is not selfish, even if it is software that you did most of the work on. Why else would one do it, other than their own self-interest. I suppose the software may improve with people getting more money to work on it, but that's almost a moot point if the improvements are proprietary. As a user of Free software, I am used to the people developing it being around to receive bug reports, make small fixes when needed, etc. if nothing more. Taking software private makes it abandonware, unless it is popular enough to attract some people willing to keep the project going, as in the case of OpenSSH (which has been very successful, due to the usefulness of the SSH to people who are good at coding.)
Anyway, for better or for worse, taking Free software private is selfish. Everyone has a certain degree of greed. I'm not saying (and I don't think Bruce is saying) that nobody should ever take software private, but I wish people wouldn't do it. They have every right to do so, but I see it as a cop-out. I know you have no problem with proprietary software, etc., I just don't like the way you seem unwilling to be realistic about your views. What I think you are saying is that it's ok for people to be selfish. Fine. Don't try to call it propaganda-busting.
> The GPL claims to be about freedom, but is actually intended to destroy programmers' livelihoods and companies.
No, it's intended to destroy the current business model of software companies. gnu.org says as much in their philosophy pages. As I see it, writing proprietary software is a bad business to get into. This view works for me because I just graduated, and haven't really committed to a career. (I want to design network hardware, or something like that. I do physics as well as CS. If I just did programming, I would want to have people pay me to write software that they needed to accomplish something, not so they could sell copies of it. That may be hard to achieve, and this whole parenthetical bit is way off topic, so I'll drop it.) Proprietary software will take long enough to go away (if it ever does) that current not a lot of career-programmers will be on the street because of the GPL. (The current economic situation has more to do with recent problems than the GPL itself.)
> The GPL does not. The "giving" forced by the GPL is not "giving" at all, because it is compulsory. It does not leave one free to give or not give.
GPLed software is a gift to like-minded programmers, and it is nothing to creators of proprietary software. It is not a gift to everyone, but it is still a gift. If you don't give gifts, you don't get gifts. Makes sense to me.
> And it has done quite well as this. Andy Hertzfeld's company is already gone
If you're trying to say that you can't build a business on selling support for GPLed software, how do you explain the fact that Cygnus consistently makes a profit. (Probably not a consistent profit, I haven't heard about their profitability the past year either, so correct me if I'm wrong.)
If you're going to reply to me, I should give you a brief outline of my views: I think proprietary software has its uses, but only in "leaf" applications. For example, an Atomic Force Microscope interface program can be proprietary without much harm. (In the case I encountered in the university physics lab a couple summers ago, a Free control program would have been more useful, because we could have make a better tool out of it with more control of the software.) In any case, the AFM control program wasn't a key part of any system other than the one made up of that software and the AFM hardware. OTOH, software like a windowing toolkit (e.g. KDE) really really should be Free. Lots of other software is built upon it, and it wouldn't be possible to use software that required proprietary components on a Free system. (This only matters to ideological purists like myself in the case where the toolkit is free enough to not require royalty payments for distributing software using it, and to allow porting to other computer architectures.) Even KDE wasn't Free enough at first, thus the GNOME project. Basically, general purpose, commonly used software (i.e. the building blocks of a computing environment) should be Free, IMHO. This is important, because it allows anyone to contribute to the evolution of the system.
It was obvious from the tone of the comment he made that he wasn't picking on Hawk, just pointing out that he hadn't dotted all his 'i's, as it were. He was pointing out that doing so is not necessary. Note that he didn't tell Hawk to go away and not come back until he knew HTML or anything. I thought he made his point quite tactfully.
BTW, I agree with his main point, especially his observation that too many lawyers generate more lawyers. The system enters "legal runaway" (once the positive feedback coefficient is greater than one).
I don't write code for Berlin, and I haven't even gotten around to trying it out... Maybe now that it's useful enough for someone to make a .deb of it, I'll give it a shot. Thus, I don't know the answer to your question about invokeNow(), etc.
:) (or on the remote machine, if the server is not on the same machine as the client).
You could make a progress bar async by having a display thread that drew the bar at a length corresponding to the current amount of progress. (so updates the the bar length that happened while the display thread was blocked would all be skipped, except the last one.)
So, yeah, the drawing thread still blocks, but at least it won't be slowing down your chess analysis on your dual proc machine
SHM is only used for put/get image stuff, AFAIK. :1 ...). If you set DISPLAY=localhost:0, the xlib will use TCP. (This is occasionally useful when hacking around if you want to run an X client from inside a chroot, where the Unix-domain sockets in /tmp aren't available).
Local communication happens over Unix sockets (/tmp/.X11-unix/X0, etc.) when DISPLAY=:0 (or
Latency is a problem here, as well as bandwidth. On a large-latency link (e.g. CA*Net3, a multi-Gb/s link between several Canadian Universities), the round trip latency for light-in-fiber alone is ~80ms. This means that the program blocks for that long whenever it asks the server to do anything. Displaying a progress bar could slow down progress significantly, unless the process was multithreaded, with a thread dedicated to communicating with the server. (This would make communication asynchrounous.)
> OpenSSH. :-)
:) I'm just trying to be realistic.
>
> 'nuff said.
OpenSSH has been (and is!) a lot more successful than I think an openGFS project would be. There is a critical need for secure remote access by far more people than there is for good distributed filesystems. There aren't as many Beowulf hackers as there are internet-using hackers. (that's a lot of coders...)
I don't think OpenGFS would reach a critical mass of developers. Of course, the most important consideration is how good the last Free version of GFS was, and how much work it is to keep up with what Sinistra does with their GFS.
I don't want to discourage anyone from working on this; I would love to be proved wrong
A short time after I found out about Unix from a CS course at university, I got a PC to use as a Linux workstation, instead of messing with MiNT on my puny Atari (mega 4 STe, ~80MB HD). I knew of MiNT, but I never used it.
Anyway, my major reason for not mentioning MiNT is that AFAIK no Atari ever shipped with MiNT installed. You had to do that yourself. My (possibly incorrect) impression of it was that, like Linux, it was best suited to the geek crowd.
I use ulimit -S -v 400000 to limit the virtual memory size of any one process to ~400MB on my machine with 512MB of RAM, and a bit of swap. This way, if a program decides to eat memory like crazy, it will be stopped before it can deprive any other processes of memory. This way, netscape dies, instead of init. :)
/etc/profile, where everyone gets it. It won't cause trouble for anyone, because those who need to can override it.
I'm writing a program to do physics simulations (soap froth systems, BTW), and if I goof up and give it command line options that would make it thrash, it effectively runs out of memory _before_ it can make the system thrash.
Since I only set a soft limit, I can jack it up without having to be root or anything. Thus, this is a good thing to put in
No sane user installs a new kernel without leaving the option of booting the old one. So the user may be pissed off because the effort of compiling and rebooting was wasted, but the system will still be bootable without resorting to that dust-encrusted floppy drive.
(if not, then that's a lesson learned... a simple boot off the RH CD will get you back in business.)
GNU parted is good. It can't move and resize at the same time, so there are some things you can't do with it, but that's in the works.
For shuffling data around, partimage is good. It is basically dump(8), but it will do ext2, FAT, and maybe some others.
I heard about a NIC that plugged into the ACSI port. (for non-Atarians, this is SCSI with s/small/Atari/. It's a similar protocol, and it supports DMA. It's the fastest read-write IO port on the ST. (the cartridge port is on the CPU data & address busses, but the non-programmable MMU bus-errors if you try to write to it. The only write path is through the address bus, reading addresses where the low bits are the data.))
What do you expect when there is a system call to run a function in supervisor mode? (m68k jargon' this is the same as "kernel mode" on processors that were designed to run Real OSes... (m68000 (found in all ST machines) didn't have a programmable MMU. 68020 supported an external one. 68030 and up had it built in.))
The GUI never got around to supporting more than one app open at the same time. (You could have some desk accessories running, though.) Basically, any game that needed a decent frame rate took over the whole computer anyway. Most commercial games came on floppies that you booted from. The base OS (in ROM) just booted the code from the floppy, so the GUI (also in ROM) never had any of its code executed. So not only can any game bring the whole thing down, most commercial games require you to bring your system down before playing them!
Sometimes trying to figure out how to tell it to do something is like playing a Sierra game, though. (give plasma to ugly dude -> you can't do that... That word is not in the Andromedan dictionary... etc.)
Simple stuff is pretty easy to tell MacOS about, but sometimes it's hard to figure out where the controls for something is. (This wouldn't be a problem after plenty of experience, though. Obviously the most important consideration is how easy it is to use for people who use it a lot, so this isn't too bad.)
Nice idea, but Intel already has conditional move, and maybe some other conditional insns. For web serving and compiling, and other integer tasks, a bunch of ARM processors might well do a good job. I wouldn't guess that it would be particularly easy to implement a very fast ARM machine, since the insn set is not simpler than the Alpha's for example, I don't think.
Anyway, it might be worth doing this, but probably not, unless it can be done cheaply.
#define X(x,y) x##y
fast, power efficient, x86 compat: choose 2. x86 doesn't pipeline easily. It's complicated to implement in a way that runs fast, as evidenced by everyone who's tried, such as Intel, AMD, or Transmeta. If you wanted power-efficient x86, you'd have to give up on pipelining and stuff, i.e. your chip would be slow, like a 386.
#define X(x,y) x##y
If you used Compaq/Digital's optimizing C compiler, you obviously found that the compile took a long time, because the compiler spends a long time scheduling the instructions for the Alpha's pipeline and exact insn execution capability. I read somewhere that the compiler actually simulates an Alpha running the code to see what it can do to make faster code. No wonder it takes a long time, and no wonder the code it generates is so good.
Compilation speed is nice, esp. when developing software, but you can usually get that by turning optimization off. When you're compiling something that will eventually use more CPU cycles than it took to compile, it's ok if the compile takes a long time, as long as the compiler does something useful with that time!
#define X(x,y) x##y