Well, I suppose 180 vs 240 isn't completely terrible, but it in no way compares to broadcast. There's a reason broadcast video was on minimum 3/4" tapes - there were actual FCC standards about what quality you could broadcast in, how much noise, how black your black could be, etc.
Good news everybody! 10 seconds of research has discovered that copyright infringment involves SENDING copyrighted material, not receiving it! Usenet is only dangerous to contributors.
I suppose that no googlites in their ivory tower has heard about the firefox trademark issue, and certainly has no idea how trivial it is to determine if a browser is firefox based despite it's silly name.
No, they insist I go from Iceweasel 3.5 to Official firefox 2.0 in order to try out their toys.
Let me think about this tradeoff for a second. Hrm. no.
except you're completely and utterly wrong and have probably never read the standard boilerplate employment agreements.
Unless you have a 'nice' employer, or you've specifically negotiated your contract, many (most? all?) standard contracts claim everything you've ever done as theirs.
I've had that clause signed away on my agreements in the past, and won't work anywhere that doesn't agree to that. Taking the money or not is immaterial, it's the employment agreement that specifies what they own.
Example:
You never work from home, ever. You have no "work" laptop, only a desktop. You don't connect between them at all. At 5pm every day you go home and quit thinking about work. The software you're creating in your free time on your own hardware at your own house? Yeah, you signed that over to them.
If I could get a monopoly on oxygen, I sure as hell could make lots of money - no matter what else I did wrong.
When you have a de-facto monopoly on life (Nobody aside from the top 0.1% get to chose their own health insurance - it's whatever their employer picks, which boils down to "what's the cheapest plan we can reasonably get away with") and rake in a 30% profit margin every year - even in a down economy... why, exactly, should you care about improving, well, anything?
Uninformed twats blathering my slashdot? It's more likely then you think!
I mean, seriously, what? WHAT? How did you even come up with this argument? You do realize that non-governmental hospitals have basically ZERO regulations on 'whoopsie!' moments. There's nothing about mixing up names - if some underpaid and overworked RN drops the wrong chart in your slot as you're being wheeled around, you die. Or lose a leg, or two. If anything, some government mandate to actually track the number of preventable medical incidents would give incentive to FIX their problems.
But keep railing on big government, child. While eating regulated food, not dying of diseases we've wiped out, using the internet links subsidized by the government, going to work via government roads, and squatting in your basement cave furiously mashing your fingers on a computer developed with government grants powered by government subsidized and regulated power.
I quit reading Stephenson because the operative word in "Snow crash" is "Crash" - the book is speeding along then WHAM hits a brick wall. All the important ends are wrapped up in a 2 page epilogue. It really reads like "Oh good I hit the pagecount they paid me for, I'm done now."
Gibson has his faults as well. His imagery is amazing, and the Decking of Neuromancer fame is firmly enshrined in conciousness - but when he starts writing AI poetry towards the end the whole thing just falls apart. Or maybe I didn't do enough drugs to really enjoy the ending...
Personally, I'd love to see someone do a decent adaptation of Foundation, or an I, Robot that wasn't just a name drop. Asimov wrote so many amazing works, and they're sorely underrepresented in film.
Performance is meaningless when your program is trivially exploited by hostile input. The problem with strcat is it assumes every pointer is in a unique address space with infinite ram. Since there's no such thing on any processor in existence, the shortcomings become rapidly clear.
The issue is the assumption that a single pointer to the first character of a string gives you all the information that you need - this is horribly broken. You have to have at minimum two pieces of information - where it's located, and how large the buffer allocated is. Knowing how 'full' it is currently is a nice optimization. It's not possible to represent those three things with a single char * pointer, so all the string functions are inherently broken.
Don't use null-terminated strings for text processing in your application under any circumstances. I don't care how smart you are - if you're smart enough to get it "right" you're going to be writing your own string library anyway. Get a decent C string library, or migrate to C++ and use std::string, or use another higher-level language. When you need to use a system/library call that needs a char *, you can get that from your String. (foo.c_str(), etc)
... unless that's not actually the motion in question. It certainly has a section VI "Class Action Allegations", with the class of plaintifs being "Those who were subjected to illegal and flawed investigations, were named in sham lawsuits or were threatened with suit by Defendants for file-sharing, downloading or other similar activities, who have not engaged in violation of copyright laws"
Do you have to file something else to proceed as a class action suit and is that what's sealed?
The document has only been linked 5 times in this discussion, is it really so hard to read it?
The answer is yes, it's the same stuff we all know has been happening, and they probably want it sealed so it doesn't interfere with other pending lawsuits.
18 claims of relief, tresspass to chattels, computer fraud and abuse, RICO, invasion of privacy, fraud, unlawful trade practices, false light (Libel, I think), defamation, abuse of legal process, wrongful initiation of civil proceedings, civil conspiracy, state RICO, intentional infliction of emotional distress, negligence.
109 pages of repeating the same assertions - Mediasentry is an unlicensed investigator, RIAA knew their claim was bogus, RIAA knew the name and address of the real person who shared the files, RIAA publicly claimed that she was downloading and sharing âoeShake That Ass Bitchâ, âoeDope Noseâ, âoeDie Motherfucker Dieâ, âoeBullet in the Headâ, âoeFuck yâ(TM)all Hoesâ, âoeNigger Fuckerâ, and âoeI Stab Peopleâ despite knowing these claims were not true, threatening to depose her 8 year old daughter, calling her daughter's school under false pretenses to contact her, implying that her conduct was aiding and abetting drug runners and terrorists, refusing to dismiss the false suit even after proof she was not doing the filesharing, and using federal courts as a threat to extort tens of thousands of dollars from people who couldn't afford to defend themselves against the false allegations.
I.E. nothing we on slashdot didn't already know.
Are the number of basically duplicate claims standard? For instance, the invasive drive scans MediaSentry was doing are claimed under Computer Fraud and Abuse as well as Tresspass to Chattels.
Oh yeah and I guess read "allegedly" to everything I said there. You know, with scare quotes, just like they do when (allegedly) publicly defaming people.
you're right on everything but prioritizing. Without prioritizing VoIP sucks, so you've effectively killed off all internet telephony startups. I have no problem with my "evil" monopoly cable provider prioritizing my voip calls above my bulk downloads - 10 seconds longer on a 4gb download vs not being able to make a phone call boy is that a difficult choice to make.
I've got to disagree here - it's a bad network design that treats bulk downloads the same as as time-critical packets. Your VoIP call should absolutely be prioritized above my bittorrent download - and it's easy to show why. In a congestion situation, proper QoS means that downloads may take a little longer - but VOIP STILL WORKS AT ALL.
Now, throttling torrents to modem speeds is wrong, but traffic shaping isn't the same thing as queuing priority. And honestly, unless your neighbor is running a call center, your bittorrent client is only competing with other big downloaders.
Streaming video (not video conferencing) is interesting as well - unlike VoIP it's not as time sensitive, you can have a 10 second buffer without significantly degrading the call. As long as you can sustain an average bandwidth within a given buffer-size window, your streaming video still works. It's a tough call, because it is time sensitive, but not AS time sensitive as VoIP is, and you have to distinguish between them.
Of course, you also shouldn't oversell your bandwidth 100:1 or worse...
Someone explain to me why you would _EVER_ have a long-run cable with crimped ends? Hint- DON'T DO IT. Punch that shit down, ALWAYS. Then use a commercial cat6 patch cable to go to the device.
It's a lot easier to get the punchdown right - the tools can be had for under $100 and cat5e punchdown blocks aren't much more. If you're going to an individual/pair of runs you can box it and put them behind a nice keystone faceplate.
You're still not getting it. What people are bitching about is the catistrophic data loss due to committing the metadata change before the data. If neither hits the disk, you get the old copy. If both hit the disk, you get the new copy. It's the middle that's the hairy case - if you commit the data first, then the metadata, everything is fine. If you commit the metadata first THEN the data, it means you're pointing at garbage if it crashes in a two minute perod - AND it's already deleted the previous file.
This promotes a simple 'regression of one change on a complete crash' to a 'wipe out of all configuration data'. THAT is the problem.
POSIX being "unspecified" isn't good enough to absolutely guarantee dataloss on a system crash. If I don't fsync my changes, they may not go through, but if they don't go through they shouldn't destroy something else! That's just bad design.
Generally you put explosives in children's toys. That does the most long-term damage in injuries (the young live longer with their amputated legs) and morale (parents freak right the fuck out).
KDE doesn't care that you may have to change your settings again. KDE cares that when given two files, one which was guaranteed to be on disk, and one that might be on disk, the kernel perfers to instead give you a zero byte file after a crash.
The fact there's a 5 second window in EXT3 is a bug too, there should be ZERO TIME where the rename is committed to disk but the blocks are not. That's a faulty design by arrogant filesystem designers hiding behind "POSIX SAYS IT'S UNDEFINED SO I DEFINE IT THIS WAY."
There's absolutly zero reason to ever use a journaled filesystem if it's DESIGNED to trash data like this - why? Because if the only alternative is to sync every write, why have the fucking journal at all?
The state on disk and state in RAM need to be considered seperately, and every write needs to leave the disk in a sane state. And no, deleting a file and replacing it with a zero-byte file, _THEN_ eventually getting around to filling in that file? That's not a sane state. Forcing every application to write sync? Dude, if you thought atime updates slowed down a filesystem...
The problem IS NOT that the update was lost. The problem is that when you do a write/rename sequence, ext4 happily reorders that to rename/wait 2 minutes/write. That means ext4 is deleting the 'old' copy of data then waiting two minutes to write the new.
And that's absurd. There's absolutely ZERO reason that rename (and the destruction of the original file) should happen on disk before the data is committed. It's just a horribly bad design decision that needs to be fixed. I don't care if it is "in the spec", filesystems have had such horrible fsync performance for so long that you can't blithely ignore the fact that userspace is doing write/rename to preserve files. ESPECIALLY when you STILL have the fsync performance issue!
This is absolutely NOT a userland issue. If every application that updates a file has to use fsync or EXT4 will happily delete the old and new copies, well, then you might as well just mount the partition sync then. It's a ridiculous requirement.
I'm sorry, what exactly is so horrible about holding off the metadata write until after the data is on disk? The fact that the filesystem CAN reorder the transactions and stay within spec doesn't mean it SHOULD. It's pretty fucking stupid to write 'this data is here' into the log before, you know, actually putting the data there. And that includes blocking other updates - you may in memory rename the file over the original, but on disk the dirent shouldn't be overwritten until the the replacement metadata is written, which shouldn't be written until the data is committed to disk.
And hell, that's the exact reason why fsync in userspace DOESN'T work - you can't "just write one file", it has to flush ALL remaining metadata, in a bad seek-order. Hans and his "dancing trees" were supposed to do something about this, but reiserfs is one of the worst offenders in this regard.
Per the spec - it's entirely possible that if I do the write/rename trick on two seperate files, then the system crashes, the one I did second would have "taken", and the first would not have. THAT is where you use fsync, and it's an accepted part of design. What's not acceptable is doing a write/rename and having the filesystem trash both copies!
Slashdot Mirror
In slashdot math, 180 lines is better then 480.
Well, I suppose 180 vs 240 isn't completely terrible, but it in no way compares to broadcast. There's a reason broadcast video was on minimum 3/4" tapes - there were actual FCC standards about what quality you could broadcast in, how much noise, how black your black could be, etc.
Good news everybody! 10 seconds of research has discovered that copyright infringment involves SENDING copyrighted material, not receiving it! Usenet is only dangerous to contributors.
Leech away!
I suppose that no googlites in their ivory tower has heard about the firefox trademark issue, and certainly has no idea how trivial it is to determine if a browser is firefox based despite it's silly name.
No, they insist I go from Iceweasel 3.5 to Official firefox 2.0 in order to try out their toys.
Let me think about this tradeoff for a second. Hrm. no.
except you're completely and utterly wrong and have probably never read the standard boilerplate employment agreements.
Unless you have a 'nice' employer, or you've specifically negotiated your contract, many (most? all?) standard contracts claim everything you've ever done as theirs.
I've had that clause signed away on my agreements in the past, and won't work anywhere that doesn't agree to that. Taking the money or not is immaterial, it's the employment agreement that specifies what they own.
Example:
You never work from home, ever. You have no "work" laptop, only a desktop. You don't connect between them at all. At 5pm every day you go home and quit thinking about work. The software you're creating in your free time on your own hardware at your own house? Yeah, you signed that over to them.
The $1250 is a red herring.
If I could get a monopoly on oxygen, I sure as hell could make lots of money - no matter what else I did wrong.
When you have a de-facto monopoly on life (Nobody aside from the top 0.1% get to chose their own health insurance - it's whatever their employer picks, which boils down to "what's the cheapest plan we can reasonably get away with") and rake in a 30% profit margin every year - even in a down economy... why, exactly, should you care about improving, well, anything?
Uninformed twats blathering my slashdot? It's more likely then you think!
I mean, seriously, what? WHAT? How did you even come up with this argument? You do realize that non-governmental hospitals have basically ZERO regulations on 'whoopsie!' moments. There's nothing about mixing up names - if some underpaid and overworked RN drops the wrong chart in your slot as you're being wheeled around, you die. Or lose a leg, or two. If anything, some government mandate to actually track the number of preventable medical incidents would give incentive to FIX their problems.
But keep railing on big government, child. While eating regulated food, not dying of diseases we've wiped out, using the internet links subsidized by the government, going to work via government roads, and squatting in your basement cave furiously mashing your fingers on a computer developed with government grants powered by government subsidized and regulated power.
Obviously we'd fuck up healthcare too.
I quit reading Stephenson because the operative word in "Snow crash" is "Crash" - the book is speeding along then WHAM hits a brick wall. All the important ends are wrapped up in a 2 page epilogue. It really reads like "Oh good I hit the pagecount they paid me for, I'm done now."
Gibson has his faults as well. His imagery is amazing, and the Decking of Neuromancer fame is firmly enshrined in conciousness - but when he starts writing AI poetry towards the end the whole thing just falls apart. Or maybe I didn't do enough drugs to really enjoy the ending...
Personally, I'd love to see someone do a decent adaptation of Foundation, or an I, Robot that wasn't just a name drop. Asimov wrote so many amazing works, and they're sorely underrepresented in film.
(In case it's not obvious, this comment was directed at gangien, not hyc)
Performance is meaningless when your program is trivially exploited by hostile input. The problem with strcat is it assumes every pointer is in a unique address space with infinite ram. Since there's no such thing on any processor in existence, the shortcomings become rapidly clear.
The issue is the assumption that a single pointer to the first character of a string gives you all the information that you need - this is horribly broken. You have to have at minimum two pieces of information - where it's located, and how large the buffer allocated is. Knowing how 'full' it is currently is a nice optimization. It's not possible to represent those three things with a single char * pointer, so all the string functions are inherently broken.
Don't use null-terminated strings for text processing in your application under any circumstances. I don't care how smart you are - if you're smart enough to get it "right" you're going to be writing your own string library anyway. Get a decent C string library, or migrate to C++ and use std::string, or use another higher-level language. When you need to use a system/library call that needs a char *, you can get that from your String. (foo.c_str(), etc)
... unless that's not actually the motion in question. It certainly has a section VI "Class Action Allegations", with the class of plaintifs being "Those who were subjected to illegal and flawed investigations, were named in sham lawsuits or were threatened with suit by Defendants for file-sharing, downloading or other similar activities, who have not engaged in violation of copyright laws"
Do you have to file something else to proceed as a class action suit and is that what's sealed?
The document has only been linked 5 times in this discussion, is it really so hard to read it?
The answer is yes, it's the same stuff we all know has been happening, and they probably want it sealed so it doesn't interfere with other pending lawsuits.
18 claims of relief, tresspass to chattels, computer fraud and abuse, RICO, invasion of privacy, fraud, unlawful trade practices, false light (Libel, I think), defamation, abuse of legal process, wrongful initiation of civil proceedings, civil conspiracy, state RICO, intentional infliction of emotional distress, negligence.
109 pages of repeating the same assertions - Mediasentry is an unlicensed investigator, RIAA knew their claim was bogus, RIAA knew the name and address of the real person who shared the files, RIAA publicly claimed that she was downloading and sharing âoeShake That Ass Bitchâ, âoeDope Noseâ, âoeDie Motherfucker Dieâ, âoeBullet in the Headâ, âoeFuck yâ(TM)all Hoesâ, âoeNigger Fuckerâ, and âoeI Stab Peopleâ despite knowing these claims were not true, threatening to depose her 8 year old daughter, calling her daughter's school under false pretenses to contact her, implying that her conduct was aiding and abetting drug runners and terrorists, refusing to dismiss the false suit even after proof she was not doing the filesharing, and using federal courts as a threat to extort tens of thousands of dollars from people who couldn't afford to defend themselves against the false allegations.
I.E. nothing we on slashdot didn't already know.
Are the number of basically duplicate claims standard? For instance, the invasive drive scans MediaSentry was doing are claimed under Computer Fraud and Abuse as well as Tresspass to Chattels.
Oh yeah and I guess read "allegedly" to everything I said there. You know, with scare quotes, just like they do when (allegedly) publicly defaming people.
you're right on everything but prioritizing. Without prioritizing VoIP sucks, so you've effectively killed off all internet telephony startups. I have no problem with my "evil" monopoly cable provider prioritizing my voip calls above my bulk downloads - 10 seconds longer on a 4gb download vs not being able to make a phone call boy is that a difficult choice to make.
I've got to disagree here - it's a bad network design that treats bulk downloads the same as as time-critical packets. Your VoIP call should absolutely be prioritized above my bittorrent download - and it's easy to show why. In a congestion situation, proper QoS means that downloads may take a little longer - but VOIP STILL WORKS AT ALL.
Now, throttling torrents to modem speeds is wrong, but traffic shaping isn't the same thing as queuing priority. And honestly, unless your neighbor is running a call center, your bittorrent client is only competing with other big downloaders.
Streaming video (not video conferencing) is interesting as well - unlike VoIP it's not as time sensitive, you can have a 10 second buffer without significantly degrading the call. As long as you can sustain an average bandwidth within a given buffer-size window, your streaming video still works. It's a tough call, because it is time sensitive, but not AS time sensitive as VoIP is, and you have to distinguish between them.
Of course, you also shouldn't oversell your bandwidth 100:1 or worse...
UNLIMITED INTERNET ACCESS $49.99*
<font size=-99>*usage charges apply</font>
Someone explain to me why you would _EVER_ have a long-run cable with crimped ends? Hint- DON'T DO IT. Punch that shit down, ALWAYS. Then use a commercial cat6 patch cable to go to the device.
It's a lot easier to get the punchdown right - the tools can be had for under $100 and cat5e punchdown blocks aren't much more. If you're going to an individual/pair of runs you can box it and put them behind a nice keystone faceplate.
You're still not getting it. What people are bitching about is the catistrophic data loss due to committing the metadata change before the data. If neither hits the disk, you get the old copy. If both hit the disk, you get the new copy. It's the middle that's the hairy case - if you commit the data first, then the metadata, everything is fine. If you commit the metadata first THEN the data, it means you're pointing at garbage if it crashes in a two minute perod - AND it's already deleted the previous file.
This promotes a simple 'regression of one change on a complete crash' to a 'wipe out of all configuration data'. THAT is the problem.
POSIX being "unspecified" isn't good enough to absolutely guarantee dataloss on a system crash. If I don't fsync my changes, they may not go through, but if they don't go through they shouldn't destroy something else! That's just bad design.
Rene Magritte might.
Generally you put explosives in children's toys. That does the most long-term damage in injuries (the young live longer with their amputated legs) and morale (parents freak right the fuck out).
I challenge you to try it yourself before spewing opinions out your gaping anus.
Take a linux box with say ext2 or ext3. Run on it a while, get used to how it works.
Now, mount -o sync your partitions. Enjoy the blazing speed of your modern hardware!
KDE doesn't care that you may have to change your settings again.
KDE cares that when given two files, one which was guaranteed to be on disk, and one that might be on disk, the kernel perfers to instead give you a zero byte file after a crash.
The fact there's a 5 second window in EXT3 is a bug too, there should be ZERO TIME where the rename is committed to disk but the blocks are not. That's a faulty design by arrogant filesystem designers hiding behind "POSIX SAYS IT'S UNDEFINED SO I DEFINE IT THIS WAY."
There's absolutly zero reason to ever use a journaled filesystem if it's DESIGNED to trash data like this - why? Because if the only alternative is to sync every write, why have the fucking journal at all?
The state on disk and state in RAM need to be considered seperately, and every write needs to leave the disk in a sane state. And no, deleting a file and replacing it with a zero-byte file, _THEN_ eventually getting around to filling in that file? That's not a sane state. Forcing every application to write sync? Dude, if you thought atime updates slowed down a filesystem...
The problem IS NOT that the update was lost. The problem is that when you do a write/rename sequence, ext4 happily reorders that to rename/wait 2 minutes/write. That means ext4 is deleting the 'old' copy of data then waiting two minutes to write the new.
And that's absurd. There's absolutely ZERO reason that rename (and the destruction of the original file) should happen on disk before the data is committed. It's just a horribly bad design decision that needs to be fixed. I don't care if it is "in the spec", filesystems have had such horrible fsync performance for so long that you can't blithely ignore the fact that userspace is doing write/rename to preserve files. ESPECIALLY when you STILL have the fsync performance issue!
This is absolutely NOT a userland issue. If every application that updates a file has to use fsync or EXT4 will happily delete the old and new copies, well, then you might as well just mount the partition sync then. It's a ridiculous requirement.
You're misunderstanding. Atomicity means what you SHOULD lose is the fact that you changed your desktop background, not everything.
Either old Or New - atomicity. NOT random data, NOT gone entirely.
Yup, your math is right, it pretty much cancels out.
World class runner can move at 4.5m/s, .22 round may move at say 400m/s.
so the formula is atan(running * time / distance)
atan(4.5 * .25 / 100) and
atan(4.5 * 2.5 / 1000)
Which are identical, obviously. So your lead would be identical for both shots.
I'm sorry, what exactly is so horrible about holding off the metadata write until after the data is on disk? The fact that the filesystem CAN reorder the transactions and stay within spec doesn't mean it SHOULD. It's pretty fucking stupid to write 'this data is here' into the log before, you know, actually putting the data there. And that includes blocking other updates - you may in memory rename the file over the original, but on disk the dirent shouldn't be overwritten until the the replacement metadata is written, which shouldn't be written until the data is committed to disk.
And hell, that's the exact reason why fsync in userspace DOESN'T work - you can't "just write one file", it has to flush ALL remaining metadata, in a bad seek-order. Hans and his "dancing trees" were supposed to do something about this, but reiserfs is one of the worst offenders in this regard.
Per the spec - it's entirely possible that if I do the write/rename trick on two seperate files, then the system crashes, the one I did second would have "taken", and the first would not have. THAT is where you use fsync, and it's an accepted part of design. What's not acceptable is doing a write/rename and having the filesystem trash both copies!