Sheesh, I was half asleep when I posted before
on
YETI@Home
·
· Score: 1
I've never written FFT code myself, but I know what they are (being a physics + CS student and all.) I have done fourier transforms (not numerical, just integrating sin(n*omega_nought) * f(x,t) (and cos(), and for different n to get the frequency space representation of the wave.) algebraicly) by hand on functions our physics prof for Oscillations and Waves assigned.
You are correct that my argument was a bit fluffy, hehe:) The no particular order bit was dumb. It probably just runs through the whole data set many times. (It might well do other interesting stuff, and/or have working copies of the data which are that size again, though. I think that's what I was thinking when I posted my first msg.) If it just worked on the first bit of the data, then the next bit, etc. then the kernel would swap out the parts it wasn't using. This doesn't happen, so I assume it dances over its memory space fairly thoroughly. You do agree that the algorithms they use require this much memory to process data that size, don't you? #define X(x,y) x##y
You're right about the network load. You forgot about server load, though. Assuming the server is running a traditional ftpd (e.g. wu-ftpd), it forks a heavyweight process (as oppose to a lightweight proc == thread) and execs/usr/sbin/in.ftpd for every new ftp session. Again, this only happens once at startup. But don't forget that ftp is a very stateful protocol. Each invocation of the daemon uses a significant amount of RAM, and you have one process for each connection. (err, I guess apache has one proc per connection, but I don't think it touches much RAM after forking.) IMHO, it would be a Good Thing if ftp disappeared as a protocol for anonymous file transfers of a single file. (ftp has the advantage of being able to tell the client how the symlinks go, which is handy for mirroring Debian with its binary-all directory and binary-{alpha,i386,etc.} directories with symlinks back to binary-all for platform independent stuff, like docs and scripts.) Don't argue that FTP can restart transfers. HTTP can do that too. Read the man page for wget sometime:)
In case you're wondering, there was a nice debate about this on our LUG mailing list. Take a look at the archives of the discussion: http://nslug.ns.ca/pipermail/nslug/19 99-June/thread.html (the discussion started with an ftp URI for metalab, and I suggested that http is better/faster #define X(x,y) x##y
Linux is good for developers because it is hard to write a program to crash the system (unless run with root privs, of course. and I'm ignoring root exploits _in other programs_ here.) (AFAIK, no programs are known to be able to crash the linux kernel (i.e. make it panic.)) Since you can't even write a program to crash the system if you are trying to, you aren't too likely to break the system with the program you're developing, no matter what you're doing. (unless you're writing a kernel module or a rootly program to tweak kernel parms.) I haven't said anything about the possibilities of DOS by getting daemons to die, but it is very, very hard to get sshd to die, so at least the admin can fix it:) (If any of the above is false, I would like to hear about it. Please back up any assertions with facts/URIs to docs I should have read before spouting off...)
I don't know how WinNT or Win2k is at this, but I know the Mac sux rocks as a development platform. Macs have two huge problems. Memory allocation (the app needs more memory than it is allocated. "Sorry, we got lazy when we wrote that part of the OS. but hey, the GUI looks nice, doesn't it:)", and cooperative multitasking. I've seen a guy lose work he had started to type in MS Word on a Mac (a ppc with at least MacOS 7.5, IIRC.). He ran NCSA telnet (*puke*, use NiftyTelnet SSH) to check on some facts. It turned out that the network connection didn't work, (cable unplugged, plugging in didn't help), and NCSA telnet hung indefinitely. After a couple hours, he rebooted. While the computer was waiting for the TCP connection, _you couldn't switch to another program_, _you couldn't cancel the connection attempt_, you couldn't do anything. For all intents and purposes, the machine was locked hard. Don't remember whether the Command+power key worked, but nothing would unwedge this dumb OS. Not very impressive, especially compared to the UltraSPARC workstations in the next lab over, with GNU emacs LaTeX:) Anyway, the co-operative multitasking is why it's such a pain in the ass for d.net to write clients for the mac.
This issue bit me a while ago. MacOS can't even multitask an MP3 player, it seems. On an iMac (original model, G3 @ 233, 32MB RAM, crap speakers), running an mp3 player (don't remember which program specifically) in the background while doing a bit of spreadsheet and graphing (really light workload!), there was a noticable pause between typing something and having it appear on the screen. Stopping the MP3 player fixed this. So, you say that MacOS can run 5 programs at once? It can't even run 2! You can have lots of programs open, but not getting the CPU. If you have a few programs which all want the CPU, your Mac won't like it. (I sincerely hope that MacOS X mostly fixes this. There's certainly hope, since somebody here said it's based on FreeBSD 3.2, which multitasks quite nicely (even though it isn't Linux, you/. raving idiots).:) It is an abomination to have MacOS running on the hardware it does, and wasting it so horribly. AFAIK, the free Unix operating systems which work on PPC hardware all kick butt. BeOS is pretty sweet too. (I've only seen the x86 version, though. What impresses me most is that they have the sense to include GNU bash 2.0 and GCC, plus posix utils:) #define X(x,y) x##y
I didn't see any link. This is all there is: (User Info) http://stats.distributed.net/rc5-64/psummary.php3? id=226557 Eyyy, I shouldn't be such a troll. But I was really pissed when I disassembled that client. There should be a warning on it. See the link. D.nnrw,rpne! mfspr r3, pc / lvxl v0, 0, r3 / li r0, 16 / stvxl v0, r3, r0 [ Reply to This | Parent ]
BTW, I've disassembled the client, and it didn't piss me off. I've also taken a look at the source for the cruncher part of the client. (not that I got very far with it... It might help if I looked up the algorithm before trying to figure out the super-optimized code which runs it...:) #define X(x,y) x##y
err, no, wait. It's a moose; it's a squirrel; it's a yeti!:) (Boris, moose and squirrel have triggered yeti detector again. No-goodniks:) #define X(x,y) x##y
Re:YETI as the so-called "missing link"
on
YETI@Home
·
· Score: 1
They attempt to paint us as bumpkins, as idiots, as hillbillies.
and
hose that believe in and study that foul theory [evolution] are to be despised, hated, and imprisoned
Your hypocrisy is dripping out around the edges. You might want to look into that.
Why do you think that the fact that a "missing link" has not been found disproves the theory of evolution? (Yes, it is only a theory. Nobody knows _for sure_ that it is right.) If I lost my watch, then I can't find it, does that mean my watch never existed? no, didn't think. I suggest that your argument parallels the watch argument, and thus is faulty. If you can find positive evidence for Creation, then that's great. Has it ever occured to you that even if evolution has happened on earth, the universe had to be created somehow. I believe that God created the universe, and lets things run from there. I haven't made up my mind on the question of miracles, whether He set up the universe so things would happen the way they are ahead of time, or whether He goes in and tweaks things every now and then. However it works, God is cool:)
also note that God could have created yetis. The truth condition of the evolution theory has no relevance on the search for yetis. The situation now would be the same if the universe had popped into place 5 minutes ago with everything having their current positions, velocities, and internal states (including your and my minds, and the yeti lurking in my back yard:), as it would be if the universe had evolved to this point from the big bang, or if God had at some point in history created everything (with positions and velocities such that the universe would eventually reach the exact state it's in now, of course)
You'll probably be upset to know that I am a volunteer Sunday school teacher every week at my church. (and no I don't try to tell the kids about my ideas about the universe and stuff. It's a bit deep for grade 6 students, so I just stick to the "normal" stuff. (most of the time:))
BTW, I would be more impressed if you had logged in to post that. We can tell you stand up for your views when you don't bother to even identify yourself. You probably realized some idiot would moderate you down, but still. I definitely don't think you should get moderated down for pointing out that creation is a possibility. I suppose you did deserve to be moderated down, since we all know Christ teaches us not to hate or despise anyone, even those who don't agree with us. I think you honestly believe the theory of Creation, and that's ok with me. Argue your point rationally, and I'll believe what I want to believe. I'll let you believe what you want to believe, but I am of course free to propose rational arguments why your position is wrong. (I won't do that here, since I'm sure you've seen many rational arguments for evolution. I choose to merely refute your argument against evolution.) #define X(x,y) x##y
Re:YETI@Home hits the nail on the head
on
YETI@Home
·
· Score: 1
It uses a lot of memory because it is doing FFTs and stuff on big arrays. The algorithm it runs fundamentally needs to touch all that memory in no specific order. (they could have written it to not use much RAM and do all the array accesses to disk. But then it would take about 50 000 times longer to complete a work unit.:( I run d.net clients on my computers because it only touches about 50-100 kB of memory while it is crunching, only touching more when it saves/loads packets. I wish it did something useful instead of try to brute force RC5, but it is easy to run and takes care of itself well. (I don't have time to babysit it on lots of compters, which is why I spend my hours reading slashdot... right. I think I figured it out. I hopefully can convince myself that I don't need to be glued to my tube every day.) #define X(x,y) x##y
Corel doesn't represent Linux. Most people don't run Corel Linux. (All Right Thinking people run Debian GNU/Linux, by my definition of "Right Thinking. Slack is ok, too.:)
It is rare that something like this isn't the case, though. Perhaps that's how us linux users get away with it. Yay, more distros, so we can't accept blame for security holes found in any given distro:) (actually, there are way too many distros.) #define X(x,y) x##y
Linux has been and always will be free. (If someone tries to make it non-free, I will personally shove RMS (that's right, beard and all) down that person's throat.)
As for the price of distributions, they seem to want to make some money off it by selling free support and/or a book with the distro. There is nothing wrong with that, but I don't like the impression it gives to new users. (i.e. the impression that linux is expensive just like windoze. (it isn't because you can legally copy it, see below, and because you don't have to buy any more stuff to do useful work.))
Of course, the best way to get into linux is to find a local LUG, since you can ask questions, and get extremely useful info about how to set up linux to work with the local ISPs unfriendly setup. Even better, you can take your computer to a meeting and have some expert hackers work on getting your (random hardware X) supported, etc. Also, you can get someone with a fast 'net connection and a burner to make a CD of the distro of your choice for $2 a CD. (even ones like Stampede, which is available only by download. I don't know if Debian is being sold or not, I heard something about a retail Debian. (and I _don't_ mean Corel's linux.)) #define X(x,y) x##y
disclaimer: I haven't done my homework on this, so I might be wrong.
Does the redhat problem let people without accounts on the machine get root? no. Therefore, it is a much less severe problem, IMHO. Presumably, you at least have some idea who your users are, and if not then you are probably a big site with professional admins who will be very suspicious about _every_ suid binary, etc. RedHat comes with a whole lot of SUID shit (yes, shit. normally I don't curse stuff, but rh has way too much suid stuff. I haven't looked at how complicated any of the programs are, but just the number of them seems risky.)
Did RedHat themselves write the buggy software? If so, or if the author has a fix which RH didn't get, then thumbs down to RH.
Does RedHat brag about how secure their distribution is? no. Does Micros~1? yes. Now who do we laugh at when old unfixed security problems are found in each.
Also, keep in mind that RedHat is not the only linux distro. Debian rocks. Debian gets fixed fast if there is a problem, and you don't have to dig through security reports if you don't want to. (you should, but you are more or less safe if you apt-get update; apt-get upgrade every now and then, AFAIK. You can read the changelog in/usr/{share,}/doc/package and see what changed, too.) #define X(x,y) x##y
Will people stop comparing Win2k to linux kernel x.x.x? No bugs have been found in the Windows 2000 kernel (yet.) Bugs have been found in the daemons^H^H^H^H^H^H^Hservices it starts by default, but that happens to all linux distributions.
However, I don't really think that is much excuse for Micros~1. Everything they release has been written in recent years, when being careful about buffer overflows is a well known programming concept. Software which is based on 10 year old or more code (like sendmail and wu-ftpd, I think) at least have the excuse that they were written before most people had reason to think about security. (of course, there isn't really much excuse to run them, give the existence of new MTAs written during the age of security (as it were:), like exim. ftpd replacements exist too.)
So, Micros~1 windows 2000 is like a distribution with some buggy programs and some configuration errors in stuff which runs in the default system. This is unacceptable, given that you are paying money for Win2k. It doesn't bug me too much to have a problem with a linux distro, because the problems get noticed and fixed in the next release of the distro. Micros~1 will be selling win2k CDs which come set up wrong for a long time. Redhat probably sells almost no CDs of rh6.0, and rh6.1 has most/all of the known security problems fixed. This is different from shipping a rh6.0 cd with a cd which upgrades it to 6.1, since it is easy to not bother doing the upgrade, especially for newbies who were overwhelmed enough by the install! Debian, of course, is the best for this. You install off some old CDs, then you apt-get upgrade and all the fixes/new versions of stuff gets installed. It's so easy even a newbie should be able to manage it. EVEN NEWBIES WHO DON'T READ CERT OR BUGTRAQ WILL GET FIXES INSTALLED easily. This is very important.
Some people have commented that every system needs to have a competent admin who reads security warnings and stuff, so it is ok to have lots of stuff enabled by default. This is all well and good, as long as linux or win2k is only used on company servers. Linux is used by people with cable modems who don't really have a clue (some people clue in after a while, but they didn't know enough when they first installed.) Even for a good admin, it is much easier to not have to figure out what is already going on, and to be able to say, "I want mail, web, and ftp, so I'll install the packages for that, then enable it in the config files", than to say, "gee this machine seems to be running a web server already. I wonder what J. Random Hacker on the 'net can get off my machine right now?". Having to know about everything there is and then portscan your machine to see what you have just seems like a really silly arrangement to me. But remember, it should be possible for people who are just learning to install linux without worrying about getting cracked into. (and without having that happen without them knowing, let alone worrying, about the possibility!!!).
So, given that win2k is targeted at everyone, not just servers (I think), Micros~1 looks really dumb. Joe Newbie has no idea he is running insecure.exe as a service. Well, I've gone on long enough. I hope that made sense, but I'm sure my ideas jumped around faster than I could type, so I probably screwed up somewhere. Hope it makes some sense:) #define X(x,y) x##y
Oh yeah, and did you notice that the default text for the 3d text screensave is "OpenGL"? This is not a coincidence. Micros~1 has (AFAIK good) support for OpenGL. It is a standard on the Windoze platform and elsewhere. (having grown out of SGI's SGIGL, IIRC). #define X(x,y) x##y
I pity you for being stuck with so much Micros~1 stuff lying around. I've got a PC here with no direct X on it. It's running Debian woody, and it hasn't even had a DOS partition for months (and it's going to stay that way.) #define X(x,y) x##y
Yes, when you mark something as Offtopic, that's a negative point. You can fix the bug of the posts going to -1 by logging in before you post, so they only go down from 1 to 0.
Assuming you meant find | xargs rm -rfn, the find | xargs is redundant. rm -r is enough by itself to handle the recursion.
Another thing: Getting used to using rm -rfn is a REALLY REALLY BAD idea. What happens when you help a friend who has an account on a Unix box which has a (non-GNU) version of rm which recursively deletes its arguments, then prints: "-n: unkown option". Guess it sucks to be you then. (granted, this is unlikely since most rm implementations don't do anything if they don't like their command line, but if you get used to putting -n as the _last_ option on the line, and going back to delete it later, then programs which use POSIX getline (and so consider all args after the first non-option arg to be non-options), then you are screwed. #define X(x,y) x##y
I think you meant to say dd if=/dev/zero of=file count= you could probably replace with $(POSIXLY_CORRECT=1 du file | cut -f1), but that only works if the file is not sparse. anyway..., you get the point. #define X(x,y) x##y
Everyone is making a big deal about the x86 compatibility. What I want to know is how well it runs code for other processors (which is the whole point of doing everything in software, right?) It would be cool if it could run SPARC or Alpha binaries, because those chips (esp. Alpha) have much better floating point units than x86 processors. (yes I know about K7s. they're still limited by the x86 arch.). Will the Crusoe have a PPC mode, so I can run d.net's G4 Altivec cruncher? hehehe.:-)
Besides the speed factor, x86s use 80bit max floating point numbers. SPARCs can use 128bit floating point. (You get this with gcc's long double type.)
Even more interesting, what is the Crusoe's "native" instruction format like? Is there a gcc that generates native code for it that runs faster? How many registers does it have, or does it even have registers? (maybe just a bit of normal RAM which is as fast as registers, but even less special than the general purpose regs on most RISC cpus.)
I guess all of this will be answered in about 1 minute because the transmeta webpage is supposed to go live around now! #define X(x,y) x##y
Doesn't APM already do this? There's some setting in my AWARD bios which says something about CPU throttling. Is how is this different? (other than the fact that the change is triggered by power source, not system load). #define X(x,y) x##y
The English police will be coming for you both shortly. You are both guilty of high treason against the English language, included the capital offence of misuse of an ellipsis.
It's not surprising that there are a heck of a lot of howtos about ppp. 90% (warning, previous number is random) who install linux had to set up ppp. It is easy for windoze people to get ppp going (or it should be), but the ISP world apparently revolves around them, so everything is set up for windoze. With Linux, many people have an experience which wasn't quite the same as the one in the doc they read, so _being the helpful type who write stuff for other people_ they write down how to set up ppp.
The other thing is the Micros~1 has everything centralized so there isn't duplication of instructions. (not as much anyway.) Hrmmm, often it seems that not even a single copy of documentation on something is available.
BTW, I consider RH6.0's installer "easy and straightforward." same for mandrake, because it's the same installer. It took me 20 minutes to install linux on a spare partition on a friend's computer and get it to detect the NIC and connect to the internet through his ADSL connection. I didn't get it to handle his SBLive! very well. (that was before the recent developments in drivers, though). #define X(x,y) x##y
Anyone who flames you for that is not worth listening to.
Recompiling the kernel _is_ fun. I love those menus where you get to choose from all the cool stuff that Linux can do:) The critical thing to do is make a bootdisk so if (when) you screw up and make lilo not able to boot, you can get into linux and run lilo before rebooting again:)
good stuff to look at:/proc/interrupts,/proc/dma,/dev/sndstat. lspci. lsmod. good luck. #define X(x,y) x##y
They used an i820 mobo and RDRAM for the P3, but just normal 100MHz SDRAM for the K7. I'd like to see a result from a benchmark that didn't need much memory bandwidth. Also, the 133MHz bus of the fastest P3 will help it a lot when all it has to do is stuff bytes into an accelerator card. It might be that you can't buy RDRAM K7 mobos, in which case AMD needs to get some made. #define X(x,y) x##y
Slashdot Mirror
I've never written FFT code myself, but I know what they are (being a physics + CS student and all.) I have done fourier transforms (not numerical, just integrating sin(n*omega_nought) * f(x,t) (and cos(), and for different n to get the frequency space representation of the wave.) algebraicly) by hand on functions our physics prof for Oscillations and Waves assigned.
:) The no particular order bit was dumb. It probably just runs through the whole data set many times. (It might well do other interesting stuff, and/or have working copies of the data which are that size again, though. I think that's what I was thinking when I posted my first msg.) If it just worked on the first bit of the data, then the next bit, etc. then the kernel would swap out the parts it wasn't using. This doesn't happen, so I assume it dances over its memory space fairly thoroughly. You do agree that the algorithms they use require this much memory to process data that size, don't you?
You are correct that my argument was a bit fluffy, hehe
#define X(x,y) x##y
You're right about the network load. You forgot about server load, though. Assuming the server is running a traditional ftpd (e.g. wu-ftpd), it forks a heavyweight process (as oppose to a lightweight proc == thread) and execs /usr/sbin/in.ftpd for every new ftp session. Again, this only happens once at startup. But don't forget that ftp is a very stateful protocol. Each invocation of the daemon uses a significant amount of RAM, and you have one process for each connection. (err, I guess apache has one proc per connection, but I don't think it touches much RAM after forking.) IMHO, it would be a Good Thing if ftp disappeared as a protocol for anonymous file transfers of a single file. (ftp has the advantage of being able to tell the client how the symlinks go, which is handy for mirroring Debian with its binary-all directory and binary-{alpha,i386,etc.} directories with symlinks back to binary-all for platform independent stuff, like docs and scripts.) Don't argue that FTP can restart transfers. :)
9 99-June/thread.html (the discussion started with an ftp URI for metalab, and I suggested that http is better/faster
HTTP can do that too. Read the man page for wget sometime
In case you're wondering, there was a nice debate about this on our LUG mailing list. Take a look at the archives of the discussion:
http://nslug.ns.ca/pipermail/nslug/1
#define X(x,y) x##y
Linux is good for developers because it is hard to write a program to crash the system (unless run with root privs, of course. and I'm ignoring root exploits _in other programs_ here.) (AFAIK, no programs are known to be able to crash the linux kernel (i.e. make it panic.))
Since you can't even write a program to crash the system if you are trying to, you aren't too likely to break the system with the program you're developing, no matter what you're doing. (unless you're writing a kernel module or a rootly program to tweak kernel parms.)
I haven't said anything about the possibilities of DOS by getting daemons to die, but it is very, very hard to get sshd to die, so at least the admin can fix it
(If any of the above is false, I would like to hear about it. Please back up any assertions with facts/URIs to docs I should have read before spouting off...)
I don't know how WinNT or Win2k is at this, but I know the Mac sux rocks as a development platform.
Macs have two huge problems. Memory allocation (the app needs more memory than it is allocated. "Sorry, we got lazy when we wrote that part of the OS. but hey, the GUI looks nice, doesn't it:)", and cooperative multitasking. I've seen a guy lose work he had started to type in MS Word on a Mac (a ppc with at least MacOS 7.5, IIRC.). He ran NCSA telnet (*puke*, use NiftyTelnet SSH) to check on some facts. It turned out that the network connection didn't work, (cable unplugged, plugging in didn't help), and NCSA telnet hung indefinitely. After a couple hours, he rebooted. While the computer was waiting for the TCP connection, _you couldn't switch to another program_, _you couldn't cancel the connection attempt_, you couldn't do anything. For all intents and purposes, the machine was locked hard. Don't remember whether the Command+power key worked, but nothing would unwedge this dumb OS. Not very impressive, especially compared to the UltraSPARC workstations in the next lab over, with GNU emacs
LaTeX
Anyway, the co-operative multitasking is why it's such a pain in the ass for d.net to write clients for the mac.
This issue bit me a while ago. MacOS can't even multitask an MP3 player, it seems. On an iMac (original model, G3 @ 233, 32MB RAM, crap speakers), running an mp3 player (don't remember which program specifically) in the background while doing a bit of spreadsheet and graphing (really light workload!), there was a noticable pause between typing something and having it appear on the screen. Stopping the MP3 player fixed this. So, you say that MacOS can run 5 programs at once? It can't even run 2! You can have lots of programs open, but not getting the CPU. If you have a few programs which all want the CPU, your Mac won't like it. (I sincerely hope that MacOS X mostly fixes this. There's certainly hope, since somebody here said it's based on FreeBSD 3.2, which multitasks quite nicely (even though it isn't Linux, you
all kick butt. BeOS is pretty sweet too. (I've only seen the x86 version, though. What impresses me most is that they have the sense to include GNU bash 2.0 and GCC, plus posix utils:)
#define X(x,y) x##y
(User Info)
http://stats.distributed.net/rc5-64/psummary.php3
Eyyy, I shouldn't be such a troll. But I was really pissed when I
disassembled that client. There should be a warning on it. See the
link.
D.nnrw
r3, r0
[ Reply to This | Parent ]
BTW, I've disassembled the client, and it didn't piss me off. I've also taken a look at the source for the cruncher part of the client. (not that I got very far with it... It might help if I looked up the algorithm before trying to figure out the super-optimized code which runs it...:)
#define X(x,y) x##y
yes.
:) (Boris, moose and squirrel have triggered yeti detector again. No-goodniks :)
err, no, wait. It's a moose; it's a squirrel; it's a yeti!
#define X(x,y) x##y
and
Your hypocrisy is dripping out around the edges. You might want to look into that.
Why do you think that the fact that a "missing link" has not been found disproves the theory of evolution? (Yes, it is only a theory. Nobody knows _for sure_ that it is right.) If I lost my watch, then I can't find it, does that mean my watch never existed? no, didn't think. I suggest that your argument parallels the watch argument, and thus is faulty. If you can find positive evidence for Creation, then that's great. Has it ever occured to you that even if evolution has happened on earth, the universe had to be created somehow. I believe that God created the universe, and lets things run from there. I haven't made up my mind on the question of miracles, whether He set up the universe so things would happen the way they are ahead of time, or whether He goes in and tweaks things every now and then. However it works, God is cool :)
also note that God could have created yetis. The truth condition of the evolution theory has no relevance on the search for yetis. The situation now would be the same if the universe had popped into place 5 minutes ago with everything having their current positions, velocities, and internal states (including your and my minds, and the yeti lurking in my back yard :), as it would be if the universe had evolved to this point from the big bang, or if God had at some point in history created everything (with positions and velocities such that the universe would eventually reach the exact state it's in now, of course)
You'll probably be upset to know that I am a volunteer Sunday school teacher every week at my church. (and no I don't try to tell the kids about my ideas about the universe and stuff. It's a bit deep for grade 6 students, so I just stick to the "normal" stuff. (most of the time :))
BTW, I would be more impressed if you had logged in to post that. We can tell you stand up for your views when you don't bother to even identify yourself. You probably realized some idiot would moderate you down, but still. I definitely don't think you should get moderated down for pointing out that creation is a possibility. I suppose you did deserve to be moderated down, since we all know Christ teaches us not to hate or despise anyone, even those who don't agree with us. I think you honestly believe the theory of Creation, and that's ok with me. Argue your point rationally, and I'll believe what I want to believe. I'll let you believe what you want to believe, but I am of course free to propose rational arguments why your position is wrong. (I won't do that here, since I'm sure you've seen many rational arguments for evolution. I choose to merely refute your argument against evolution.)
#define X(x,y) x##y
It uses a lot of memory because it is doing FFTs and stuff on big arrays. The algorithm it runs fundamentally needs to touch all that memory in no specific order. (they could have written it to not use much RAM and do all the array accesses to disk. But then it would take about 50 000 times longer to complete a work unit. :( I run d.net clients on my computers because it only touches about 50-100 kB of memory while it is crunching, only touching more when it saves/loads packets. I wish it did something useful instead of try to brute force RC5, but it is easy to run and takes care of itself well. (I don't have time to babysit it on lots of compters, which is why I spend my hours reading slashdot... right. I think I figured it out. I hopefully can convince myself that I don't need to be glued to my tube every day.)
#define X(x,y) x##y
Corel doesn't represent Linux. Most people don't run Corel Linux. (All Right Thinking people run Debian GNU/Linux, by my definition of "Right Thinking. Slack is ok, too. :)
:) (actually, there are way too many distros.)
It is rare that something like this isn't the case, though. Perhaps that's how us linux users get away with it. Yay, more distros, so we can't accept blame for security holes found in any given distro
#define X(x,y) x##y
As for the price of distributions, they seem to want to make some money off it by selling free support and/or a book with the distro. There is nothing wrong with that, but I don't like the impression it gives to new users. (i.e. the impression that linux is expensive just like windoze. (it isn't because you can legally copy it, see below, and because you don't have to buy any more stuff to do useful work.))
Of course, the best way to get into linux is to find a local LUG, since you can ask questions, and get extremely useful info about how to set up linux to work with the local ISPs unfriendly setup. Even better, you can take your computer to a meeting and have some expert hackers work on getting your (random hardware X) supported, etc. Also, you can get someone with a fast 'net connection and a burner to make a CD of the distro of your choice for $2 a CD. (even ones like Stampede, which is available only by download. I don't know if Debian is being sold or not, I heard something about a retail Debian. (and I _don't_ mean Corel's linux.))
#define X(x,y) x##y
disclaimer: I haven't done my homework on this, so I might be wrong.
/usr/{share,}/doc/package and see what changed, too.)
Does the redhat problem let people without accounts on the machine get root? no.
Therefore, it is a much less severe problem, IMHO. Presumably, you at least have some idea who your users are, and if not then you are probably a big site with professional admins who will be very suspicious about _every_ suid binary, etc. RedHat comes with a whole lot of SUID shit (yes, shit. normally I don't curse stuff, but rh has way too much suid stuff. I haven't looked at how complicated any of the programs are, but just the number of them seems risky.)
Did RedHat themselves write the buggy software? If so, or if the author has a fix which RH didn't get, then thumbs down to RH.
Does RedHat brag about how secure their distribution is? no. Does Micros~1? yes. Now who do we laugh at when old unfixed security problems are found in each.
Also, keep in mind that RedHat is not the only linux distro. Debian rocks. Debian gets fixed fast if there is a problem, and you don't have to dig through security reports if you don't want to. (you should, but you are more or less safe if you apt-get update; apt-get upgrade every now and then, AFAIK. You can read the changelog in
#define X(x,y) x##y
Will people stop comparing Win2k to linux kernel x.x.x? No bugs have been found in the Windows 2000 kernel (yet.) Bugs have been found in the daemons^H^H^H^H^H^H^Hservices it starts by default, but that happens to all linux distributions.
However, I don't really think that is much excuse for Micros~1. Everything they release has been written in recent years, when being careful about buffer overflows is a well known programming concept. Software which is based on 10 year old or more code (like sendmail and wu-ftpd, I think) at least have the excuse that they were written before most people had reason to think about security. (of course, there isn't really much excuse to run them, give the existence of new MTAs written during the age of security (as it were:), like exim. ftpd replacements exist too.)
So, Micros~1 windows 2000 is like a distribution with some buggy programs and some configuration errors in stuff which runs in the default system. This is unacceptable, given that you are paying money for Win2k. It doesn't bug me too much to have a problem with a linux distro, because the problems get noticed and fixed in the next release of the distro. Micros~1 will be selling win2k CDs which come set up wrong for a long time. Redhat probably sells almost no CDs of rh6.0, and rh6.1 has most/all of the known security problems fixed. This is different from shipping a rh6.0 cd with a cd which upgrades it to 6.1, since it is easy to not bother doing the upgrade, especially for newbies who were overwhelmed enough by the install! Debian, of course, is the best for this. You install off some old CDs, then you apt-get upgrade and all the fixes/new versions of stuff gets installed. It's so easy even a newbie should be able to manage it. EVEN NEWBIES WHO DON'T READ CERT OR BUGTRAQ WILL GET FIXES INSTALLED
easily. This is very important.
Some people have commented that every system needs to have a competent admin who reads security warnings and stuff, so it is ok to have lots of stuff enabled by default. This is all well and good, as long as linux or win2k is only used on company servers. Linux is used by people with cable modems who don't really have a clue (some people clue in after a while, but they didn't know enough when they first installed.) Even for a good admin, it is much easier to not have to figure out what is already going on, and to be able to say, "I want mail, web, and ftp, so I'll install the packages for that, then enable it in the config files", than to say, "gee this machine seems to be running a web server already. I wonder what J. Random Hacker on the 'net can get off my machine right now?". Having to know about everything there is and then portscan your machine to see what you have just seems like a really silly arrangement to me. But remember, it should be possible for people who are just learning to install
linux without worrying about getting cracked into. (and without having that happen without them knowing, let alone worrying, about the possibility!!!).
So, given that win2k is targeted at everyone, not just servers (I think), Micros~1 looks really dumb. Joe Newbie has no idea he is running insecure.exe as a service. Well, I've gone on long enough. I hope that made sense, but I'm sure my ideas jumped around faster than I could type, so I probably screwed up somewhere. Hope it makes some sense:)
#define X(x,y) x##y
Oh yeah, and did you notice that the default text for the 3d text screensave is "OpenGL"? This is not a coincidence. Micros~1 has (AFAIK good) support for OpenGL. It is a standard on the Windoze platform and elsewhere. (having grown out of SGI's SGIGL, IIRC).
#define X(x,y) x##y
I pity you for being stuck with so much Micros~1 stuff lying around. I've got a PC here with no direct X on it. It's running Debian woody, and it hasn't even had a DOS partition for months (and it's going to stay that way.)
#define X(x,y) x##y
Yes, when you mark something as Offtopic, that's a negative point. You can fix the bug of the posts going to -1 by logging in before you post, so they only go down from 1 to 0.
Sheesh, some people.
#define X(x,y) x##y
Another thing: Getting used to using rm -rfn is a REALLY REALLY BAD idea. What happens when you help a friend who has an account on a Unix box which has a (non-GNU) version of rm which recursively deletes its arguments, then prints: "-n: unkown option". Guess it sucks to be you then. (granted, this is unlikely since most rm implementations don't do anything if they don't like their command line, but if you get used to putting -n as the _last_ option on the line, and going back to delete it later, then programs which use POSIX getline (and so consider all args after the first non-option arg to be non-options), then you are screwed.
#define X(x,y) x##y
That was supposed to be ... count=<blocks in file>
dd
and you can replace <blocks in file> with $(...)
The "post as plain text" option drops angle brackets. Grrrrrr. Bug report send to cmdrtaco.
#define X(x,y) x##y
I think you meant to say dd if=/dev/zero of=file count=
you could probably replace with $(POSIXLY_CORRECT=1 du file | cut -f1), but that only works if the file is not sparse.
anyway..., you get the point.
#define X(x,y) x##y
Everyone is making a big deal about the x86 compatibility. What I want to know is how well it runs code for other processors (which is the whole point of doing everything in software, right?) It would be cool if it could run SPARC or Alpha binaries, because those chips (esp. Alpha) have much better floating point units than x86 processors. (yes I know about K7s. they're still limited by the x86 arch.). Will the Crusoe have a PPC mode, so I can run d.net's G4 Altivec cruncher? hehehe. :-)
Besides the speed factor, x86s use 80bit max floating point numbers. SPARCs can use 128bit floating point. (You get this with gcc's long double type.)
Even more interesting, what is the Crusoe's "native" instruction format like? Is there a gcc that generates native code for it that runs faster? How many registers does it have, or does it even have registers? (maybe just a bit of normal RAM which is as fast as registers, but even less special than the general purpose regs on most RISC cpus.)
I guess all of this will be answered in about 1 minute because the transmeta webpage is supposed to go live around now!
#define X(x,y) x##y
doh, I didn't mean to post this twice. Sorry.
#define X(x,y) x##y
Doesn't APM already do this? There's some setting in my AWARD bios which says something about CPU throttling. Is how is this different? (other than the fact that the change is triggered by power source, not system load).
#define X(x,y) x##y
The English police will be coming for you both shortly. You are both guilty of high treason against the English language, included the capital offence of misuse of an ellipsis.
And DON'T do it AGAIN. -- Cleese in Life of Brian
#define X(x,y) x##y
It's not surprising that there are a heck of a lot of howtos about ppp. 90% (warning, previous number is random) who install linux had to set up ppp. It is easy for windoze people to get ppp going (or it should be), but the ISP world apparently revolves around them, so everything is set up for windoze. With Linux, many people have an experience which wasn't quite the same as the one in the doc they read, so _being the helpful type who write stuff for other people_ they write down how to set up ppp.
The other thing is the Micros~1 has everything centralized so there isn't duplication of instructions. (not as much anyway.) Hrmmm, often it seems that not even a single copy of documentation on something is available.
BTW, I consider RH6.0's installer "easy and straightforward." same for mandrake, because it's the same installer. It took me 20 minutes to install linux on a spare partition on a friend's computer and get it to detect the NIC and connect to the internet through his ADSL connection. I didn't get it to handle his SBLive! very well. (that was before the recent developments in drivers, though).
#define X(x,y) x##y
Anyone who flames you for that is not worth listening to.
:) The critical thing to do is make a bootdisk so if (when) you screw up and make lilo not able to boot, you can get into linux and run lilo before rebooting again :)
/proc/interrupts, /proc/dma, /dev/sndstat. lspci. lsmod.
Recompiling the kernel _is_ fun. I love those menus where you get to choose from all the cool stuff that Linux can do
good stuff to look at:
good luck.
#define X(x,y) x##y
sorry to hear that, for your sake. I pity you.
#define X(x,y) x##y
They used an i820 mobo and RDRAM for the P3, but just normal 100MHz SDRAM for the K7. I'd like to see a result from a benchmark that didn't need much memory bandwidth. Also, the 133MHz bus of the fastest P3 will help it a lot when all it has to do is stuff bytes into an accelerator card.
It might be that you can't buy RDRAM K7 mobos, in which case AMD needs to get some made.
#define X(x,y) x##y