I've got a question for you, since you seem to like JIT so much...
Why the hell doesn't Java cache the compiled code between program executions? Program portability is nice, but 98% of us desktop users install an app and only run the installed app on a single hardware system. After the first time JIT compilation occurs, every subsequent JIT is a waste until some hardware parameter changes that would affect performance. Would it really be that hard to have a system-wide cache that the JVM could check when I run Ant so it wouldn't have to JIT it every time?
Did you actually read the page you linked to? They drank water with LSD dissolved in it, then smoked the "banana joints". Wonder which one caused the psychadelic experiences?
The whole point of X.509 certificates is that your name is attached. It's meant for verification of identity, not anonymous transactions. If you took away the name component, there would be no security because the other party has no guarentee that you are who you say you are. It'd be like having a driver's license with no name or picture, just a little notice saying that "someone" is licensed to drive. If you need some secure way of making anonymous transactions, I'd imagine a one-way auth system like SSL would work better.
Agreed, they should've told the user how to open up the right port(s), then close them after setup is done. Let's face it, though. SP2 is still in the Release Candidates, and this is a beta. I'd be very surprised if they didn't fix this by final release.
Memory controllers have been on the motherboard side for a long time, at least on x86. In fact, some of the recent chips have actually integrated them on to the processor. Having the memory controller on the RAM stick would make RAM sticks very much more expensive to produce, and the margins are already razor-thin.
"A Jabber server running at each ISP would make IM more distributed."
I like my IM centralized, and I can't see any benefits to distributing it like that. With my centralized IM, I can connect to one or two networks and get 99% of my friends. If Jabber became popular, would I have to connect to the ISP-run server of everyone on my contact list?
Even if they addressed that via a forwarding mechanism or something, they're just following the footsteps of another tech that already does what they want: IRC. It just seems like a whole lot of work and technology for no usable benefit.
I use CVS WineX, and you couldn't be more wrong. It runs DirectX absolutely fine. The proprietary parts are the CD Copy Protection code and their slick game installer/manager. If you bypass the first with a crack and can manage to install the games manually, then CVS WineX works just as well as the real deal. Stop with the FUD about "binary-only, proprietary, closed-source DirectX shared objects".
I doubt you really know what's going on under the hood of either DDD or VS.NET. If you have a huge program, you usually have separate source files which are compiled into object code which is then conglomerated into a library of code: DLL on windows,.so on Linux. When you change code while debugging, the current file is recompiled, reinserted and the execution point is re-engaged from the beginning of the program to where you left off. Now, VS.NET may have optimizations and better integration to do this, but you can obtain similar behavior with DDD and be similarly productive.
Believe it or not, VS.Net lets you edit your file, recompile parts of it, and insert the code into memory *without* resetting the debugger. If you're single-stepping through your code and you see something that you need to change, you can change it without starting your program from the beginning. It takes the level of recompilation down from object file level to function or statement level. Obviously, this isn't appropriate in every situation, and using it carelessly can get confusing. But it's helpful to me, and I use it pretty often. Here's a
MSDN page that explains it a lot better than I most likely am. It's the VC6 page on E&C, but it applies to VS.Net too.
Try this in VS.NET: while editing a line of your program, quickly throw an exit() statement maybe 20 lines above. Then when you reset your execution point, what does the program do? Does it exit? How do you debug when effects of previous logic statements change prior to the execution point? Reasonably, VS.net would re-execute these statements to get a picture of how the program would be even if ran from the beginning.
Naturally, it wouldn't exit immediately. It can't retroactively change what your program did. If I wanted to do that, I'd just restart the debugger precisely how you suggested. The code that's changed is run when the execution point reaches it again, just like any other code. It's only useful if you change code that either hasn't run yet or will be run again. It won't go back and rerun a function on all the previous inputs it has received, but it will run the newly-compiled function on succesive calls. E&C is a supplemental debugging tool, not a replacement for recompiling and restarting the debugger.
It's not a FUD statement. I'm not preaching VS.Net over anything else, just asking how I can get a VS.Net feature in my Linux dev environment. If I could use that feature with GCC and Vi or equivalent, I'd be more comfortable working in Linux, and I'd use it a little more than I currently do.
The specific feature that I really like in VS.Net is that I can be debugging, figure out what's wrong, change one line in my source, and tell VS.Net to compile that line and patch it into the running program so I can continue debugging without a full recompile. What you suggested is functionally close, but it still involves a full recompile and reload of the debugger. I was hoping for a solution that would let me recompile only the changed code bit and continue debugging without reloading. Again, it's not something absolutely essential to my work, but I'd be more productive and happier if I could do it on Linux in addition to Windows. Sorry about the misunderstanding.
Windows Longhorn will have a much improved CLI, even with a shell scripting language. I've played with the alphas a little, and it's looking to be much improved. Even so, WinXP adds a lot of command line goodness that most people don't know about. I'm too lazy to find a link, but a Google query should turn info up.
Strange coincidence, aterm breaks *my* command lines weird. If I type a line that wraps, it won't move that line up and start a new line, but it will just move back to the start of the line and overwrite. Then, when I backspace over the newline, it copies the line up one line and deletes the last character every time I press backspace. Really weird problem.
...without re-compiling or restarting the process!
gdb
Maybe I'm smoking the same crack that grandparent was, but how in the hell do you setup "Edit and Continue" with GCC/GDB or GCC/DDD? Bonus points if it works in Eclipse too. I could move a lot of my development over to my Linux boxes if I could get that working.
For reference, I run VS.Net 2003 Pro on a 1.8 Ghz P4 Mobile with 384mb RAM.
It runs fine. It takes a few seconds to start up, but that's not really an issue cause I rarely close it. Install time really isn't an issue unless you happen to be schizophrenically installing and uninstalling it every time you use it. It's not laggy during usage in any way, and compiles are speedy. If you prefer to use SciTE, that's nice, but it's nowhere near VS.Net for features. They're not even in the same class. That's like saying that Maya is bloated shitware, and you prefer to use Blender. If you wanna compare VS.Net to anything, compare it to Eclipse with CDT and JDT, KDevelop, or Anjuta. All of them are monsters, with the possible exception of Anjuta (I haven't used it enough).
Great post, just wanted to mention that 'ls -1 ~/somedir/*.c | less' does the same thing as 'ls -1 ~/somedir/ | grep "\.c$" | less", but with one less pipe and no regexps.
I've found that KDE and Mozilla (among other apps) are *much* more dependent on hard drive speed than CPU. Get a WD Raptor or similar, and both start up pretty much instantly;)
New York has such a law. Most states don't. One site I found claims that 23 states have such age ranges. Some states, such as California, have civil penalties for sex with a minor when the ages of the victim and perpetrator are similar. The catch is that the district attorney can press charges for these penalties even if all parties involved don't want to. After doing this extra research, I'd like to modify my statement to include the phrase "in some states".
The US laws are generally even more fucked up than most realize. If I was 17, and I took a picture of my 17 year old girlfriend nude, that would be considered child abuse and the picture kiddie porn. If I had sex with my 17 year old girlfriend, that's perfectly legal. However, the day I turn 18 it becomes statutory rape (as long as she is still l7). Please note that this age varies from state-to-state. Most states define the age of consent to be between 16 and 18, depending on the state. The discriminating age for child pornography (18) does not vary.
Of course, the whole concept is bullshit. When you turn 18, you don't get magically smarter. You aren't endowed with some kind of better judgement that lets you decide whether or not you want someone to take pictures of yourself naked. In fact, even after you're 18, you still can't make some decisions regarding this subject. If I take a picture of myself naked when I'm 17, then send it to someone when I'm 19, I'm distributing child pornography! I'm not even sure of the legality of *looking* at that picture after I'm 18. Does that make any sense whatsoever?
" Why not delete infected files and restore from backup? That's what backups are for."
Backups are great, but they are out of date the second you change the file after a backup is made. If I can press a button and there's a 1% chance that I'll get the file back good-as-new, I'll take that chance. Best case, it works and I go back to work. Worst case, it doesn't work and I try plan #2.
"I think it's pretty silly to add features to a product that will only work 1% of the time."
In any other situation, I'd agree with this statement. However, data recovery is a situation where your product can't afford to ignore a situation where any recovery is possible. True, you don't rely only on a virus program's recovery, similar to how you don't rely only on backups or proactive virus scanning. It's just another thing to try when restoring from backup would result in data loss. Yes, it's a bad situation, and it's much better to avoid. However, it happens in the real world. Try explaining to your boss that there was something else you could've tried to get that critical data back, but you didn't bother since there was only a 1% chance that it would've worked. 1% chance is better than 0%, which is what you'd have without that feature. If you're in that situation, then the thing you did wrong was using that software in the first place.
Plenty of viruses don't completely mangle the files they infect. Even if they did, I'm amazed they can take such a view of my data. If a virus somehow manages to infect an important file, I want every single fucking option they can give me to fix that file. I don't care if there's only a 1% chance it'll work, I wanna try it. Just because a feature is "very seldom" useful doesn't mean that you should exclude it on principle. To the people whose systems are affected, it's very useful indeed. That kind of philosophy means that even if it's simple to remove a particular virus from a file, I get to pop open my hex editor and salvage what I can. More likely, I'll choose to use a virus scanner that actually does what I expect it to be able to do.
As a sibling mentioned, Matlab and Mathematica are nice, although Maxima and Octave might be more accessible. I'd also recommend Expression Calculator. It's the best non-CAS Win32 calculator I've ever seen. I'd easily say that it's better than my TI-86.
Re:Real-world examples of tangible benefits
on
Mono Beta 2 Released
·
· Score: 1
Mods: Please mod Miguel up so the stupid troll moderation goes away. He's not a troll, nor is is comment. He's the leader of the project in question, so pretty much anything he has to say in this conversation is on-topic and untrollish.
To whoever moderated him troll: Get a life. Just because you don't agree with someone's philosophy or don't like the project they represent doesn't mean you should mod them down. If you really have some problem with Miguel, reply and let it be known to the world. Maybe you should try doing something productive with your time instead of attacking others from behind the veil of anonymity.
Miguel: Great work on MonoDevelop. It's so good to have an IDE specifically designed for the language you're working on, instead of having to use java-mode in XEmacs...
Slashdot Mirror
I've got a question for you, since you seem to like JIT so much...
Why the hell doesn't Java cache the compiled code between program executions? Program portability is nice, but 98% of us desktop users install an app and only run the installed app on a single hardware system. After the first time JIT compilation occurs, every subsequent JIT is a waste until some hardware parameter changes that would affect performance. Would it really be that hard to have a system-wide cache that the JVM could check when I run Ant so it wouldn't have to JIT it every time?
Did you actually read the page you linked to? They drank water with LSD dissolved in it, then smoked the "banana joints". Wonder which one caused the psychadelic experiences?
The whole point of X.509 certificates is that your name is attached. It's meant for verification of identity, not anonymous transactions. If you took away the name component, there would be no security because the other party has no guarentee that you are who you say you are. It'd be like having a driver's license with no name or picture, just a little notice saying that "someone" is licensed to drive. If you need some secure way of making anonymous transactions, I'd imagine a one-way auth system like SSL would work better.
"I heard this being called C-Pound in the states"
:)
Nope, we pronounce it "C-Sharp" just like the rest of the (English-speaking) world
Agreed, they should've told the user how to open up the right port(s), then close them after setup is done. Let's face it, though. SP2 is still in the Release Candidates, and this is a beta. I'd be very surprised if they didn't fix this by final release.
Memory controllers have been on the motherboard side for a long time, at least on x86. In fact, some of the recent chips have actually integrated them on to the processor. Having the memory controller on the RAM stick would make RAM sticks very much more expensive to produce, and the margins are already razor-thin.
"A Jabber server running at each ISP would make IM more distributed."
I like my IM centralized, and I can't see any benefits to distributing it like that. With my centralized IM, I can connect to one or two networks and get 99% of my friends. If Jabber became popular, would I have to connect to the ISP-run server of everyone on my contact list?
Even if they addressed that via a forwarding mechanism or something, they're just following the footsteps of another tech that already does what they want: IRC. It just seems like a whole lot of work and technology for no usable benefit.
I use CVS WineX, and you couldn't be more wrong. It runs DirectX absolutely fine. The proprietary parts are the CD Copy Protection code and their slick game installer/manager. If you bypass the first with a crack and can manage to install the games manually, then CVS WineX works just as well as the real deal. Stop with the FUD about "binary-only, proprietary, closed-source DirectX shared objects".
Believe it or not, VS.Net lets you edit your file, recompile parts of it, and insert the code into memory *without* resetting the debugger. If you're single-stepping through your code and you see something that you need to change, you can change it without starting your program from the beginning. It takes the level of recompilation down from object file level to function or statement level. Obviously, this isn't appropriate in every situation, and using it carelessly can get confusing. But it's helpful to me, and I use it pretty often. Here's a MSDN page that explains it a lot better than I most likely am. It's the VC6 page on E&C, but it applies to VS.Net too.
Try this in VS.NET: while editing a line of your program, quickly throw an exit() statement maybe 20 lines above. Then when you reset your execution point, what does the program do? Does it exit? How do you debug when effects of previous logic statements change prior to the execution point? Reasonably, VS.net would re-execute these statements to get a picture of how the program would be even if ran from the beginning.
Naturally, it wouldn't exit immediately. It can't retroactively change what your program did. If I wanted to do that, I'd just restart the debugger precisely how you suggested. The code that's changed is run when the execution point reaches it again, just like any other code. It's only useful if you change code that either hasn't run yet or will be run again. It won't go back and rerun a function on all the previous inputs it has received, but it will run the newly-compiled function on succesive calls. E&C is a supplemental debugging tool, not a replacement for recompiling and restarting the debugger.
It's not a FUD statement. I'm not preaching VS.Net over anything else, just asking how I can get a VS.Net feature in my Linux dev environment. If I could use that feature with GCC and Vi or equivalent, I'd be more comfortable working in Linux, and I'd use it a little more than I currently do.
The specific feature that I really like in VS.Net is that I can be debugging, figure out what's wrong, change one line in my source, and tell VS.Net to compile that line and patch it into the running program so I can continue debugging without a full recompile. What you suggested is functionally close, but it still involves a full recompile and reload of the debugger. I was hoping for a solution that would let me recompile only the changed code bit and continue debugging without reloading. Again, it's not something absolutely essential to my work, but I'd be more productive and happier if I could do it on Linux in addition to Windows. Sorry about the misunderstanding.
Thank you oh so much for that little bit of wisdom. I owe you my happiness for the week!
You're not entirely talking to yourself ;)
Windows Longhorn will have a much improved CLI, even with a shell scripting language. I've played with the alphas a little, and it's looking to be much improved. Even so, WinXP adds a lot of command line goodness that most people don't know about. I'm too lazy to find a link, but a Google query should turn info up.
Strange coincidence, aterm breaks *my* command lines weird. If I type a line that wraps, it won't move that line up and start a new line, but it will just move back to the start of the line and overwrite. Then, when I backspace over the newline, it copies the line up one line and deletes the last character every time I press backspace. Really weird problem.
For reference, I run VS.Net 2003 Pro on a 1.8 Ghz P4 Mobile with 384mb RAM.
It runs fine. It takes a few seconds to start up, but that's not really an issue cause I rarely close it. Install time really isn't an issue unless you happen to be schizophrenically installing and uninstalling it every time you use it. It's not laggy during usage in any way, and compiles are speedy. If you prefer to use SciTE, that's nice, but it's nowhere near VS.Net for features. They're not even in the same class. That's like saying that Maya is bloated shitware, and you prefer to use Blender. If you wanna compare VS.Net to anything, compare it to Eclipse with CDT and JDT, KDevelop, or Anjuta. All of them are monsters, with the possible exception of Anjuta (I haven't used it enough).
Great post, just wanted to mention that 'ls -1 ~/somedir/*.c | less' does the same thing as 'ls -1 ~/somedir/ | grep "\.c$" | less", but with one less pipe and no regexps.
"KDE/Mozilla run ass-slow on a 2.8ghz Xeon"
;)
I've found that KDE and Mozilla (among other apps) are *much* more dependent on hard drive speed than CPU. Get a WD Raptor or similar, and both start up pretty much instantly
New York has such a law. Most states don't. One site I found claims that 23 states have such age ranges. Some states, such as California, have civil penalties for sex with a minor when the ages of the victim and perpetrator are similar. The catch is that the district attorney can press charges for these penalties even if all parties involved don't want to. After doing this extra research, I'd like to modify my statement to include the phrase "in some states".
The US laws are generally even more fucked up than most realize. If I was 17, and I took a picture of my 17 year old girlfriend nude, that would be considered child abuse and the picture kiddie porn. If I had sex with my 17 year old girlfriend, that's perfectly legal. However, the day I turn 18 it becomes statutory rape (as long as she is still l7). Please note that this age varies from state-to-state. Most states define the age of consent to be between 16 and 18, depending on the state. The discriminating age for child pornography (18) does not vary.
Of course, the whole concept is bullshit. When you turn 18, you don't get magically smarter. You aren't endowed with some kind of better judgement that lets you decide whether or not you want someone to take pictures of yourself naked. In fact, even after you're 18, you still can't make some decisions regarding this subject. If I take a picture of myself naked when I'm 17, then send it to someone when I'm 19, I'm distributing child pornography! I'm not even sure of the legality of *looking* at that picture after I'm 18. Does that make any sense whatsoever?
" Why not delete infected files and restore from backup? That's what backups are for."
Backups are great, but they are out of date the second you change the file after a backup is made. If I can press a button and there's a 1% chance that I'll get the file back good-as-new, I'll take that chance. Best case, it works and I go back to work. Worst case, it doesn't work and I try plan #2.
"I think it's pretty silly to add features to a product that will only work 1% of the time."
In any other situation, I'd agree with this statement. However, data recovery is a situation where your product can't afford to ignore a situation where any recovery is possible. True, you don't rely only on a virus program's recovery, similar to how you don't rely only on backups or proactive virus scanning. It's just another thing to try when restoring from backup would result in data loss. Yes, it's a bad situation, and it's much better to avoid. However, it happens in the real world. Try explaining to your boss that there was something else you could've tried to get that critical data back, but you didn't bother since there was only a 1% chance that it would've worked. 1% chance is better than 0%, which is what you'd have without that feature. If you're in that situation, then the thing you did wrong was using that software in the first place.
Plenty of viruses don't completely mangle the files they infect. Even if they did, I'm amazed they can take such a view of my data. If a virus somehow manages to infect an important file, I want every single fucking option they can give me to fix that file. I don't care if there's only a 1% chance it'll work, I wanna try it. Just because a feature is "very seldom" useful doesn't mean that you should exclude it on principle. To the people whose systems are affected, it's very useful indeed. That kind of philosophy means that even if it's simple to remove a particular virus from a file, I get to pop open my hex editor and salvage what I can. More likely, I'll choose to use a virus scanner that actually does what I expect it to be able to do.
As a sibling mentioned, Matlab and Mathematica are nice, although Maxima and Octave might be more accessible. I'd also recommend Expression Calculator. It's the best non-CAS Win32 calculator I've ever seen. I'd easily say that it's better than my TI-86.
Mods: Please mod Miguel up so the stupid troll moderation goes away. He's not a troll, nor is is comment. He's the leader of the project in question, so pretty much anything he has to say in this conversation is on-topic and untrollish.
To whoever moderated him troll: Get a life. Just because you don't agree with someone's philosophy or don't like the project they represent doesn't mean you should mod them down. If you really have some problem with Miguel, reply and let it be known to the world. Maybe you should try doing something productive with your time instead of attacking others from behind the veil of anonymity.
Miguel: Great work on MonoDevelop. It's so good to have an IDE specifically designed for the language you're working on, instead of having to use java-mode in XEmacs...
An interesting question, and one that I can't answer, unfortunately. Anyone with a little more knowledge of number theory care to comment?
Yep, that's the proof using infinitesimals. Thanks.