It's nice, but actually only solves one specific way applications could guarantee integrity of their files. There's numerous other assumptions one can make of an ordered filesystem in your typical program which will not be caught by this little fix.
I still think it would be a really bad idea to use any filesystem (or database for that matter) that doesn't flush data and metadata up to the same point in time.
Though, if you could buy a new DVD movie for $1 then who would wait hours for a movie to download?
Me. However, I would not be waiting on a download. It's called planning ahead. I dislike DVD's and CD's for other reasons:
1) I find them to be archaic. Music CD's hold 75 minutes of music. They get repetitive quickly and I'd have to bring a bunch of them with me everywhere I go. The whole "Album" thing is so outdated and more of a "tradition" than a necessity these days. Instead I prefer to just put my entire collection on a few memory cards, small enough to carry around in a wallet or something.
2) I see little added value in "extras", "chapters", "interviews", "fancy menu's", "stupid repetitive background music in a loop" and "FBI warnings I cannot skip". When I want to watch a movie, that's all I want. I don't want to see "Chapter 36: Neo gets his ass handed to him", I don't want to fiddle with your non-standard menu designs, I don't want hidden easter eggs, I just want to drop down on the couch, grab my remote, select a movie to my liking and watch it... immediately. In fact, if it is a particular boring episode/movie, I may even play it back at 20% increased speed (with pitch corrected audio of course) if I even bother completely watching it at all.
3) I didn't like the myriad of Vinyl albums, tapes, video cassetes and CD's I had to buy special storage closets/holders for -- infact, I've thrown them all away by now (a small fortune that did do nothing but take up space). I'm not about to start over again with DVD, Blueray or whatever the latest craze is. Discs are a thing of the past for me and I simply cannot be bothered to even walk to the player and insert the correct disc anymore.
I believe the industry knows that you cannot stop 100% of software piracy. I don't think that's their goal.
I wouldn't be so sure:) In fact, I think that's exactly their goal and the only thing stopping them from going for it is that it would likely result in copyright law being abolished or at the very least thoroughly re-examined. They'll push it as far as they can, and they'll only stop pushing when it starts to hurt their bottom line.
You should go into politics, you have about the same level of understanding of the issues involved.
Making lots of connections is not illegal, and is in fact likely to become more and more commonplace as more and more services are developed where the combined uploading / downloading power of users is leveraged to provide decentralized and cheaper services. The only reason it is not more prevalent right now is because of retarded bandwidth restrictions on connections like ADSL. This will become a thing of the past soon enough though.
You seem to think that people use their internet connection for one thing at a time, like a microwave oven. I however run a webserver, versioning system, several SQL servers, a torrent client (yup), IMAP mail server, web mail, reverse proxy, a regular proxy, do remote backups and allow some friends to have SSH/SFTP access. Sometimes I even read slashdot.
Profiling that with some simple rules is not going to work. The webserver alone would look like a P2P program with thousands of connections a day. If there happens to be some download activity as well, I'm screwed.
The patent doesn't even have to be for sale. I certainly wouldn't be stupid enough to sell it. If some country actually thinks that it is morally or legally obliged to pay me $25-$40 for every TV produced because I thought of something first, then I'd exploit it as best as I could (well, I probably wouldn't, but there's people (hiding behind faceless cooperations) far more evil than me that would).
It's just funny that people actually think that paying an inventor/artist/producer/cooperation over an over again for the same product is actually the moral thing to do. I'd probably even agree (up to a point) if it was a real person that would receive such wealth, but a faceless cooperation? Hell no.
I see it the same way. Legal services are just an economic drain on productive society. They cannot exist without the other making them similar to leeches. Having too much resources tied up in legal technicalities will eventually be a bad thing for your economy. Legal services are also something that is pretty hard to export, so they basically only hurt yourself not your "competitors".
All of these examples however differ in just one thing. Intent. Did you intentionally break someone's nose or was it an accident?
The law therefore could simply state: intentionally breaking someone's nose is illegal.
Trying to extend the text of the law to provide a fail-safe method on how to prove intent is futile. It's not possible to establish intent without cooperation from the suspect. The most we can hope for is a good guess. I mean, I could take a swing at someone intending to stop just short of breaking their nose to give them a scare. If I stumble or the target makes a sudden move, you could have dozens of witnesses seeing me break someone's nose seemingly without cause, while it was never actually my intent.
fsync-on-rename only fixes this specific problem I fear. There's many more where that came from when you donot sync meta-data and file contents up to the same point in time. A simple example:
1) Write into some application's logfile: "user X deleted file Y" 2) Delete file Y
Step 2 is just a meta-data update and could occur before step 1 is completed. If the system crashes during these operations you could end up with file Y being deleted without a log entry showing which user deleted it. This will work fine with filesystems that seemingly execute actions in order (like ext3 in default ordered mode), but would fail on filesystems that take a more dim view of what's important.
Ext4 does need fixing. It's beyond stupidity to sync meta-data to disc more frequently than the real data that programs actually call upon a filesystem to store safely. This rename problem is just the tip of the ice-berg when data and meta-data is not synced up to the same point in time.
The problem simply is that actions taken by programs are no longer seemingly executed in order. Actions have to be separated into actions that affect meta-data, and actions that affect file contents. When you do the following actions:
1) create file 2) write data to file 3) rename/delete some other file 4) write some more data
Programs should not have to care about the fact that only actions 2 and 4 actually modify file contents. They only need to know that in case of a crash the only situations that can arise are that steps were only completed upto a certain point (so, 0 to 4 steps). Ext4 did not guarantee this and could end up with steps 1 and 3 committed but not step 2 -- the end result is that to be absolutely safe, every program in existance has to be rewritten to sync every time your applications "switches" between updating meta-data and file contents.
Unfortunately, sync does FAR more than just tell the filesystem that step 3 is dependent on step 2 -- it tells it to drop everything, flush everything to disc right now, add a mark to the journal and wait until all that has completed -- not just your data, but all data that every application on your system might be working on -- it's a sledge hammer, one that should be used with care, and RARELY in user applications.
People don't seem to realize that modern filesystems (in ordered mode) already do hundreds of actions before actually committing them by default -- they can get away with that because they guarantee one simple thing: it all still LOOKS sequential, because they guarantee that a situation can never arise where a later action was committed but an earlier action was not.
Oh come now, I use VGA to display 1600x1200 at 100 fps (on a CRT). I noticed that at this speed, the picture can become slightly blurry (only slightly mind you) when using an old cheap analog VGA cable. When run at 80 fps the picture was crisp and clear again and indistinguishable from the digital equivalent.
Do I need to explain that 1600x1200 @ 80 Hz is way more bandwidth than 1920x1080 @ 25 Hz?
C++ is just not well suited for building an IDE around. That's really the power of Java/C#, they play nice with IDE's -- mainly due to the lack of macros and conditional compilation. It allows the IDE to be far more powerful and offer refactoring options I could only dream of when I was still programming C/C++. No amount of AI is gonna solve this really, for the same reason that dynamically compiled languages will never be able to do something simple as globally renaming methods/variables without side effects.
No use with 2 GB of memory? By default it will use a max of 256 MB (limited by Java VM), and I found that to be enough for projects with 5000-10000 source files.
I donot see why someone spending a year making a movie (let's assume they "worked" 8 hours a day) should earn several orders of magnitude more money than someone who spend a year working at any other job.
I especially dislike that somehow the money earned for a movie should scale with the amount of viewers. Are they saying that if a movie that made $500 million with a population of 5 billion that should entitle then to double that if there were 10 billion people? I simply see such schemes as a huge drag on the economy with that much money flowing from the lower classes directly into the pockets of the so-called elite, for what is in essence, a one time effort.
I completely agree. It's obvious that these producers are simply making money by the bucket-load. When producing a movie they simply donot know what to do with it all, resulting in outrageously overpriced actors and special effects often with a flimsy and plot-hole ridden storyline.
300 million means little when several of the people involved are paid in millions as well.
...and since there is already enough creative work out there to last anyone a life-time, it's time copyright was abolished. It's not a scarce product, as some would have you believe.
It will be solved soon enough though. Torrents without a tracker already are possible. A new protocol could arise that does away with it altogether. Bandwidth continues to grow (making freenet feasible -- it already is feasible for things like music). Wifi-style networks as an alternative to centrally managed ISP's. And in the somewhat further future: people trading tiny Memory Cards in the 100+ terabyte range with entire collections of movies/music on them, more than they'd ever watch/listen in their lifetime.
Ext3 not fragmenting is a myth anyway. Ext3 fragments just as badly with multiple concurrent writes going on (like happens on a drive with log files or lots of relatively slow updating files -- like downloads).
To be honest, I donot see why SSD's can't use a more reasonable sector size. It seems to me that if SSD manufacturers want to try and replace hard drives, they could have known that most filesystems will use block sizes below 4-8k -- and that they do this for good reason.
Hard drives can easily handle blocks of 128 kB as well but performance suffers when using such large blocks (mostly due to inefficient cache use -- hard drives these days take about the same time to randomly read/write 512 bytes as 128 kB). It's however a bit pointless to poison your cache with 128 kB of worthless data when all you need is a 4k directory block. It's hard to keep related meta data together in a filesystem beyond the block level without some kind of background re-ordering process going on. Even then, there's often no need for "other" data when all the user wants is to query the existence of a file or the contents of a directory (leaving much of a 128 kB read wasted even under optimal conditions). This is as far as I know the primary reason no filesystem will use such large block sizes by default, despite it not taking much more time to read 128 kB vs just 512 bytes.
UI performance is for me the primary focus when deciding whether I find an OS acceptable to use or not. There's no excuse these days for reaction times to not be near instant, especially for things like context menu's, tab or window switches.
Other things like animated "sliding" menu's are just plain stupid -- when I click on a menu, I usually already have a good idea of where I'm going and will already be moving the mouse to the expected location of the item I want. This is thrown off though when a menu slides in, meaning the position of the item I want is actually changing while I'm trying to focus on selecting it. Only when the animation is done I can go there. These and similar animations are just the first thing I turn off as they actually hurt usability.
Same goes for menu's that re-order themselves, have simple and expert versions (personalized menu's), they all make the same mistake: they move the location of the item I want forcing me to actually look more closely to the screen to pick the item I want. Location is more important than fancy icons or trying to be smart and moving something I use often to a "more optimal" spot.
Slashdot Mirror
It's nice, but actually only solves one specific way applications could guarantee integrity of their files. There's numerous other assumptions one can make of an ordered filesystem in your typical program which will not be caught by this little fix.
I still think it would be a really bad idea to use any filesystem (or database for that matter) that doesn't flush data and metadata up to the same point in time.
Me. However, I would not be waiting on a download. It's called planning ahead. I dislike DVD's and CD's for other reasons:
1) I find them to be archaic. Music CD's hold 75 minutes of music. They get repetitive quickly and I'd have to bring a bunch of them with me everywhere I go. The whole "Album" thing is so outdated and more of a "tradition" than a necessity these days. Instead I prefer to just put my entire collection on a few memory cards, small enough to carry around in a wallet or something.
2) I see little added value in "extras", "chapters", "interviews", "fancy menu's", "stupid repetitive background music in a loop" and "FBI warnings I cannot skip". When I want to watch a movie, that's all I want. I don't want to see "Chapter 36: Neo gets his ass handed to him", I don't want to fiddle with your non-standard menu designs, I don't want hidden easter eggs, I just want to drop down on the couch, grab my remote, select a movie to my liking and watch it... immediately. In fact, if it is a particular boring episode/movie, I may even play it back at 20% increased speed (with pitch corrected audio of course) if I even bother completely watching it at all.
3) I didn't like the myriad of Vinyl albums, tapes, video cassetes and CD's I had to buy special storage closets/holders for -- infact, I've thrown them all away by now (a small fortune that did do nothing but take up space). I'm not about to start over again with DVD, Blueray or whatever the latest craze is. Discs are a thing of the past for me and I simply cannot be bothered to even walk to the player and insert the correct disc anymore.
I wouldn't be so sure :) In fact, I think that's exactly their goal and the only thing stopping them from going for it is that it would likely result in copyright law being abolished or at the very least thoroughly re-examined. They'll push it as far as they can, and they'll only stop pushing when it starts to hurt their bottom line.
You should go into politics, you have about the same level of understanding of the issues involved.
Making lots of connections is not illegal, and is in fact likely to become more and more commonplace as more and more services are developed where the combined uploading / downloading power of users is leveraged to provide decentralized and cheaper services. The only reason it is not more prevalent right now is because of retarded bandwidth restrictions on connections like ADSL. This will become a thing of the past soon enough though.
You seem to think that people use their internet connection for one thing at a time, like a microwave oven. I however run a webserver, versioning system, several SQL servers, a torrent client (yup), IMAP mail server, web mail, reverse proxy, a regular proxy, do remote backups and allow some friends to have SSH/SFTP access. Sometimes I even read slashdot.
Profiling that with some simple rules is not going to work. The webserver alone would look like a P2P program with thousands of connections a day. If there happens to be some download activity as well, I'm screwed.
The patent doesn't even have to be for sale. I certainly wouldn't be stupid enough to sell it. If some country actually thinks that it is morally or legally obliged to pay me $25-$40 for every TV produced because I thought of something first, then I'd exploit it as best as I could (well, I probably wouldn't, but there's people (hiding behind faceless cooperations) far more evil than me that would).
It's just funny that people actually think that paying an inventor/artist/producer/cooperation over an over again for the same product is actually the moral thing to do. I'd probably even agree (up to a point) if it was a real person that would receive such wealth, but a faceless cooperation? Hell no.
I see it the same way. Legal services are just an economic drain on productive society. They cannot exist without the other making them similar to leeches. Having too much resources tied up in legal technicalities will eventually be a bad thing for your economy. Legal services are also something that is pretty hard to export, so they basically only hurt yourself not your "competitors".
All of these examples however differ in just one thing. Intent. Did you intentionally break someone's nose or was it an accident?
The law therefore could simply state: intentionally breaking someone's nose is illegal.
Trying to extend the text of the law to provide a fail-safe method on how to prove intent is futile. It's not possible to establish intent without cooperation from the suspect. The most we can hope for is a good guess. I mean, I could take a swing at someone intending to stop just short of breaking their nose to give them a scare. If I stumble or the target makes a sudden move, you could have dozens of witnesses seeing me break someone's nose seemingly without cause, while it was never actually my intent.
fsync-on-rename only fixes this specific problem I fear. There's many more where that came from when you donot sync meta-data and file contents up to the same point in time. A simple example:
1) Write into some application's logfile: "user X deleted file Y"
2) Delete file Y
Step 2 is just a meta-data update and could occur before step 1 is completed. If the system crashes during these operations you could end up with file Y being deleted without a log entry showing which user deleted it. This will work fine with filesystems that seemingly execute actions in order (like ext3 in default ordered mode), but would fail on filesystems that take a more dim view of what's important.
Ext4 does need fixing. It's beyond stupidity to sync meta-data to disc more frequently than the real data that programs actually call upon a filesystem to store safely. This rename problem is just the tip of the ice-berg when data and meta-data is not synced up to the same point in time.
The problem simply is that actions taken by programs are no longer seemingly executed in order. Actions have to be separated into actions that affect meta-data, and actions that affect file contents. When you do the following actions:
1) create file
2) write data to file
3) rename/delete some other file
4) write some more data
Programs should not have to care about the fact that only actions 2 and 4 actually modify file contents. They only need to know that in case of a crash the only situations that can arise are that steps were only completed upto a certain point (so, 0 to 4 steps). Ext4 did not guarantee this and could end up with steps 1 and 3 committed but not step 2 -- the end result is that to be absolutely safe, every program in existance has to be rewritten to sync every time your applications "switches" between updating meta-data and file contents.
Unfortunately, sync does FAR more than just tell the filesystem that step 3 is dependent on step 2 -- it tells it to drop everything, flush everything to disc right now, add a mark to the journal and wait until all that has completed -- not just your data, but all data that every application on your system might be working on -- it's a sledge hammer, one that should be used with care, and RARELY in user applications.
People don't seem to realize that modern filesystems (in ordered mode) already do hundreds of actions before actually committing them by default -- they can get away with that because they guarantee one simple thing: it all still LOOKS sequential, because they guarantee that a situation can never arise where a later action was committed but an earlier action was not.
Yes... cause only KDE apps did this... *groan*
Oh come now, I use VGA to display 1600x1200 at 100 fps (on a CRT). I noticed that at this speed, the picture can become slightly blurry (only slightly mind you) when using an old cheap analog VGA cable. When run at 80 fps the picture was crisp and clear again and indistinguishable from the digital equivalent.
Do I need to explain that 1600x1200 @ 80 Hz is way more bandwidth than 1920x1080 @ 25 Hz?
C++ is just not well suited for building an IDE around. That's really the power of Java/C#, they play nice with IDE's -- mainly due to the lack of macros and conditional compilation. It allows the IDE to be far more powerful and offer refactoring options I could only dream of when I was still programming C/C++. No amount of AI is gonna solve this really, for the same reason that dynamically compiled languages will never be able to do something simple as globally renaming methods/variables without side effects.
No use with 2 GB of memory? By default it will use a max of 256 MB (limited by Java VM), and I found that to be enough for projects with 5000-10000 source files.
Not ditching it before it also has NoScript. I seriously couldn't care less about JavaScript performance, I donot want applications in my browser.
I see it even simpler.
I donot see why someone spending a year making a movie (let's assume they "worked" 8 hours a day) should earn several orders of magnitude more money than someone who spend a year working at any other job.
I especially dislike that somehow the money earned for a movie should scale with the amount of viewers. Are they saying that if a movie that made $500 million with a population of 5 billion that should entitle then to double that if there were 10 billion people? I simply see such schemes as a huge drag on the economy with that much money flowing from the lower classes directly into the pockets of the so-called elite, for what is in essence, a one time effort.
And miss out on three people having to buy their own copy of the game?
I completely agree. It's obvious that these producers are simply making money by the bucket-load. When producing a movie they simply donot know what to do with it all, resulting in outrageously overpriced actors and special effects often with a flimsy and plot-hole ridden storyline.
300 million means little when several of the people involved are paid in millions as well.
Wow, 20 times more entertaining... I wonder what that feels like, is it better than sex?
...and since there is already enough creative work out there to last anyone a life-time, it's time copyright was abolished. It's not a scarce product, as some would have you believe.
It will be solved soon enough though. Torrents without a tracker already are possible. A new protocol could arise that does away with it altogether. Bandwidth continues to grow (making freenet feasible -- it already is feasible for things like music). Wifi-style networks as an alternative to centrally managed ISP's. And in the somewhat further future: people trading tiny Memory Cards in the 100+ terabyte range with entire collections of movies/music on them, more than they'd ever watch/listen in their lifetime.
Ext3 not fragmenting is a myth anyway. Ext3 fragments just as badly with multiple concurrent writes going on (like happens on a drive with log files or lots of relatively slow updating files -- like downloads).
To be honest, I donot see why SSD's can't use a more reasonable sector size. It seems to me that if SSD manufacturers want to try and replace hard drives, they could have known that most filesystems will use block sizes below 4-8k -- and that they do this for good reason.
Hard drives can easily handle blocks of 128 kB as well but performance suffers when using such large blocks (mostly due to inefficient cache use -- hard drives these days take about the same time to randomly read/write 512 bytes as 128 kB). It's however a bit pointless to poison your cache with 128 kB of worthless data when all you need is a 4k directory block. It's hard to keep related meta data together in a filesystem beyond the block level without some kind of background re-ordering process going on. Even then, there's often no need for "other" data when all the user wants is to query the existence of a file or the contents of a directory (leaving much of a 128 kB read wasted even under optimal conditions). This is as far as I know the primary reason no filesystem will use such large block sizes by default, despite it not taking much more time to read 128 kB vs just 512 bytes.
I think such filters might have existed in the past already, which gave rise to leet-speak.
I'll be impressed when they start blocking content with "24" and "X" in the title.
UI performance is for me the primary focus when deciding whether I find an OS acceptable to use or not. There's no excuse these days for reaction times to not be near instant, especially for things like context menu's, tab or window switches.
Other things like animated "sliding" menu's are just plain stupid -- when I click on a menu, I usually already have a good idea of where I'm going and will already be moving the mouse to the expected location of the item I want. This is thrown off though when a menu slides in, meaning the position of the item I want is actually changing while I'm trying to focus on selecting it. Only when the animation is done I can go there. These and similar animations are just the first thing I turn off as they actually hurt usability.
Same goes for menu's that re-order themselves, have simple and expert versions (personalized menu's), they all make the same mistake: they move the location of the item I want forcing me to actually look more closely to the screen to pick the item I want. Location is more important than fancy icons or trying to be smart and moving something I use often to a "more optimal" spot.
Or try get a copy of Windows 2003 -- basically XP, but with support for PAE, allowing >4 GB or RAM.
Happily running it here myself with 8 GB, and runs anything that runs on XP.