am surprised at the amount. I figured it would be reduced to be more reasonable. My big problem with all this is the damages. $18,000 per song is 900 CD sales per song at $20 a CD.
Actually the point of statutory damages is precisely to be out of proportion to the actual damage, to prevent willful copyright infringement as a standard practice. If say you could only be fined twice the actual amount of damages, you could run a business infringing copyright if only half of the copyright owners didn't sue. And except for the extreme cases they wouldn't sue because it wouldn't be worth their time and money to do so.
For example if you made $5k in a year from selling illegal copies of Windows, it would cost far more than a $10k limit for Microsoft to sue over it. So that isn't the issue at all... 'ridiculously' high statutory damages are fine.
The issue is that penalty is completely out of proportion to the person's assets. There are two problems here 1) that it's based on number of copies instead of that copies were made and 2) it's based on a fixed amount of damages instead of based on assets.
A solution is simple... make the fine for wilful violation in addition to actual damages be 0.75% to 150% of the guilty party's assets instead of $750 to $150,000. Second, make the statutory damages be based on whether copyright infringement occurred instead of how many times. So in this case you'd have 24 songs @ $1 each or $24 plus. Then this person has say 90k in assets of their estate that's an extra $675 to $135,000 total depending on how bad the jury finds the violation was. This probably isn't possible within our system, but if you share 24 songs from different rights holders then they have to fight over that 0.75% to 150%.
Ok, that's pretty funny on both counts, but I often type in a URL into a search engine to make sure I go to the legit site and not accidentally to some "whitehouse.com" or "bankfoamerica.com" type of site.
Maybe I'm just some old timer and the anti-phishing in browsers is plenty fine... idk. But anyway, the point being that all the related searches and 'this is like' sidebars can actually help a lot. Another poster says goog does this also if you select the options to do it. So anyway I'm not trying to claim Bing invented anything (it's Microsoft...), just saying that as an the overall search engine it's on par with google if not better in some ways... definitely not '10 years behind' like the original poster said. I guess I missed the joke.
Bing gives us what Google already gave us 10 years ago. This is a major advance for Microsoft.
I think that's a little bit disingenuous. Try searching "reddit.com" of google and bing.
Google gives you a list of all results mentioning reddit.com, and a few common links into the site. That's it.
Bing gives you just the entry for reddit.com (probably what you want), and the common links into the site. There's a sidebar with related searches, "reddit nsfw", "reddit game", "twitter", etc. There's a sidebar that says similar to this site is digg, drudge report, huffington post, perez hilton. Judging by experience that's a really accurate summary of reddit.com. You can click 'show all' to see other pages that match "reddit.com".
Frankly I'm pretty impressed with bing, and I can see why google would be looking at it with a keen eye.
Also what irritates me even more than just bad results (which is understandable that search isn't going to always hone in on exactly what I'm looking for), is when the keyword isn't even in the results. Even if I put the "keyword" in quotes it does the same thing... links 'pointing to the page' contained that keyword. I think there is some type of operator:keyword to actually require the word, but who knows.
Code reviews are like testing... it's only worth doing if the quality is already high. A tester that ever has to file a bug like "crashes if run with zero args, or 5 or more args" is going to shoot somebody or just phone it in. Same with code reviews. If you have to explain to somebody why their use of a bubble sort is "not a good solution" then there is no point in doing one in the first place. It's just wasting people's time.
But on the other hand if you go to a code review and see some code implementing chmod octal and oug+-rwx,... syntax in 40 lines compiling to ~200 bytes of assembly you're going to say 'damn, that's cool' and it'll pump everybody up and they might learn a few things about binary operations and masks. You might catch some errors too, that's nice, but the main point of the review should be to learn about the code and to learn better programming. If you want to use them to catch errors, you should use testing and auditing of the code instead (if the presence of a bug is important enough to convene a meeting to look the for the mere possibility of it existing, then you need to really get serious about it).
I would assume every OS has an API to do this, but it might not be as fast as you want. Certainly on a modern Linux you can look at columns 6 and 7 in/proc/schedstat for each CPU and get a pretty good idea on CPU idleness.
That's not an "API", that's a hack workaround. Open a file, have kernel format the lines, read the lines, parse the lines, figure out how many CPUs to use, create a thread, use sched_* syscalls to assign it to a CPU,... now CPU is busy again. Whoops. Even if you're using a thread pool it's so clumsy you might as well just assign it to N threads to begin with (where N is any arbitrary number). On top of that, it only works for one program at a time in the entire system (well best case it might be described by some kind of differential equation).
As far as I know, no OS has an API to do anything close to what I suggested in the post, a way to do fine-grained parallelism well. Maybe on a mainframe... Tera MTA could do this, but it had special hardware threads built into the CPU itself.
I've done some kernel work myself, and would try this, but it's not my job. And getting something like this past Ingo...
I can guarantee you that when you're at 35k feet and the speed sensor says you're going -1 kph a human pilot isn't going to aim the plane vertical and accelerate.
The big red override button sounds pretty good to me. If only they had air brakes.
I am not sure how you think that the splitting in itself would ensure immediate scheduling.
By definition. If the OS can't schedule it immediately on 4 cores, but can on 3 cores then it can split into 3 threads. No OS I know of can do this or has an API to do this.
Lo and behold, many OSes or low-level libraries also provide readily available thread pools!
Which you have to signal, then wait for the threads to be scheduled, then wait for the 'main thread' to be signaled that they completed. And you have no idea what else is going on so you don't know how many threads to use in the first place.... which was kinda the point of my post in the first place.
Any program with a "for" or "while" loop in which the results of one iteration do not depend on the results of the previous iteration... We just need the languages not to make coding this way too painful.
It's not the languages that are the problem, it's the operating systems. Check it:
1. Start 4 threads because OS said there are 4 cores 2. Each thread starts doing N/4 iterations of the loop 3. Each thread does some kind of synchronization to say it exited. It could be just a write, but there needs to be some way to wake up the original thread. 4. The original thread blocks until all the other threads have completed, then continues operating.
The best case is that there is lots of overhead creating a new thread, and the other threads have completed before the original one so that it doesn't have to wait.
The worst case is that the new threads don't get scheduled right away. Maybe flash is using 100% CPU on one core, so one thread might not even start running until one of the others completes. Anyway you end up with several threads not running and being scheduled at some later time. The for loop ends up taking several times longer even though it's running 'in parallel'.
Now what about a case like this:
1. ask os to split thread across multiple CPUs. This would be a guarantee that the threads would run immediately on the number of CPUs returned by the call. 2. run the loops on N cores (however many the OS returns). 3. each thread 'exits' when it is done. For the last thread to exit, the OS returns to it instead of exiting. Whichever thread finished last continues.
This doesn't suffer from the threads originally not starting right away and has no real synchronization except in the OS itself. It's awesome for smallish loops. It's also impossible to do with current operating systems.
The purpose of business is to make money. Not to be a patriotic cash funnel that supports governmental pet programs. Keep viewing corporations as ATM machines and they *will* relocate to more desirable locations because there are a lot of countries out there that see the benefits of all the jobs that large companies bring.
Business use a lot of government resources. The fire marshal gets called out to check that construction is following the fire codes, roads have to be built to their buildings and maintained and expanded. Water and sewer.
If you are using shared resources then you should be paying for them, in one way or another, and we have taxes to avoid the overhead of paying and accounting for every little detail by itself. Somebody has to pay to get things done, so if you are really against taxation on idealistic grounds you must be for companies paying for each and every service they use... including hard to quantify services like police and military protection.
No, a government job is not a "good" job, it is a drain on the tax base because it generates no wealth.
Generating wealth is irrelevant unless you can also protect it. If you get paid 10k and get robbed of it all then you might as well have just done nothing. Government jobs are these sorts of jobs that protect wealth or provide the support to enable others to generate wealth. You can't make a car with only an engine, and even if you try to maximize the engine by taking off "drains" like bumpers, the outer shell, upholstery, windshield, etc because theses don't 'generate' anything that doesn't make it a 'good' car.. that actually makes it a much worse car.
2. "gzip bytes," which they define as follows: "We started with the source-code markup you can see, removed comments, removed duplicate whitespace characters, and then applied minimum GZip compression."... actually #2 seems to me like a very reasonable measure of expressiveness
I'm not so sure about that. For instance, their preprocessing effectively gives Python free blocks, by removing repeated spaces but not removing { and } from other languages. What they should do is:
1. remove all comments
That's it. You don't need to remove whitespace at all... gzip will compress 10 spaces to roughly the same as 1 space, if it's regular, but reducing 4, 8, 12 runs of spaces to 1 will affect the compression. You don't need to rename variables, because AReallyLongVariable used a dozen times will be amortized away if it is used often... it won't affect the gzip size much more than calling it 'a1'. They should also use maximum compression (that's the point of compressing it in the first place, to reduce it to the essential structure)... they should use 7zip or rar. So IMO their preprocessing of the source is creating a misleading metric.
Wolfram Alpha is for when you want to know how long it will take for two orbits around the galaxy. Also, how long it takes the solar system to orbit the galaxy is not what I'd call 'useful' information...
Nobody wants to sell their pride and joy, their little tech baby. That's not the question. The question is should they sell it.
Ultimately there are just two factors to consider:
1. How much are you really going to get paid? 2. Will it still be your pride and joy?
Both are completely situational. Do you get paid up front in addition to maybe being an employee? How much do you expect to lose by selling it? Does the company want to develop the product, or shut it down? Megacorp might fire you right away, so make sure to get a guarenteed payment of some kind that is substantial. You might be able to make more money off your code yourselves, or you might fail. Nobody here can tell give you any advice, because you gave us no information.
But I would tend to say that guaranteed money is better than potential money. Also remember human nature; people that have their ships come in and get lump sums as a rule end up blowing it. You don't know what kind of taxes you get when you get a 100k lump sum (or whatever), or how fast it's going to burn. Don't go out and blow it right away on cars and fancy houses, or decide to become a day trader!
fmpeg decodes nearly everything, I don't think I've used a binary codec in quite a long time... MPEG-2, MPEG-4, H.264,.flv, wmv9, even realvideo files, ffmpeg does it natively.
Just google around a bit, for instance a post from Oct 2008: "libavcodec has some ARMv6 optimized stuff, but nothing for ARMv7 or TMS320C64x. Looks like a good project for anyone who is familiar with those"
Look at the libavcoded source... there's way more ASM for x86 and it looks like a lot more work has been done on x86 optimizations, which is to be expected. And there are still other codecs not supported at all.
it's a revelation -- you realize quick, there's not that much on a linux desktop that relies on x86.
Linux has basically zero desktop market share, so nobody really cares if their linux programs work. They care if their Windows programs work.
I've heard there's a flash for arm so that's not missing.
"I've heard" gets you +informative? Gimme a break. Adobe doesn't have a download link for flash for ARM, and linux flash copycats are buggy(er) and slow. "Not missing" and "just as good" are completely different.
And yet I hear people talking as though ARM-based netbooks will be on the market within a year or something. What am I missing here? Is it all vaporware?
It's ridiculous. People say ARM is so much more power efficient, then forget that they have to emulate x86 because some video codec isn't written for ARM. A big reason people bought linux netbooks was to put bootleg Windows on them, to run Windows apps. Windows has like 90% of the netbook market now. ARM would have to be ten times better architecture than x86 to even make a dent in netbooks, and it's barely even 1.12 times better.
Hell, as phones get even more powerful we'll probably see the opposite... x86 invading phones. After all, how much cooler would an iPhone be if you could run crysis on it?
I designed one of these about a decade ago and did some prototyping. Since I don't seem to have the time to realize it, here are a few extra features that could be added (if i2p does not already include these).
Here's a tip for anybody thinking of implementing something like i2p, tor, freenet, etc: if the user has to do anything and if it impacts performance it's not worth doing.
What's needed is something simple and pervasive:
1) compatible with regular TCP 2) optional so it is only in effect when both the source and destination support it 3) 'weak' so that there is not much performance impact, so there is no reason to disable it
For sake of example, half of an xtea key can be sent by the SYN using the TCP options field and the other half provided by the server in the ACK. If the server doesn't return its half then no encryption takes place. The key is permuted by the data sent/received.
This simple scheme provides that anybody examining the TCP steam must have seen the first packets and must have followed the entire conversation, decrypting it along the way. There is no extra step involved that could introduce delays, and the state and CPU time is small enough to be not important to the end user or server. However, for anybody to wholesale monitor traffic, for whatever reason, it means they need a lot of expensive hardware. What ends up as 1% of your CPU time to decrypt a torrent requires a warehouse of servers to decrypt everybody's torrents.
A system like this has a huge advantage over tor, freenet, etc, in that everyday normal people can have it enabled by default, especially for open-source linux, *bsd distros. The actual anonymizing networks are worthless because only those with something to hide use them, or people who are hard-core idealists (which probably also gets you on some kind of 'watch list'). Ironically, this kind of system will raise the overall cost of monitoring to a point where tor, freenet, i2p, etc become viable.
In firefox I can go full screen, but still keep the URL bar, so I know what site I'm on. I can get a menu bar by pressing Alt once. I can put the NoScript button on the URL bar. I can even put a button to toggle fullscreen, since it's often easier to trackpad to a button than find F11 on a small non-backlit keyboard. Or firefox can drop down this UI when the cursor is at the top.
Chrome can do none of these things, even in the current beta version. The UI in many ways is restrictive and lacks many niceties. You can still like it and prefer it, that's fine, but let's be honest. Perhaps I should have said "maximize space while still being usable", but some common sense is in order here.
Chromifox makes firefox look a lot like Chrome. Chrome is a nice toy, but it's UI is pretty lacking when you want to do something like maximize screen space on a 1024x600 screen.
The other thing about linux on a netbook is: firefox 3 runs javascript like a dog, and that really shows up when you don't have a beefy machine.
It's not javascript, or not just javascript. I've seen xulrunner apps that are magnitudes faster in Windows than in Linux. Apparently the rendering in Linux is not clipping correctly, so it spends a lot of extra time drawing hidden elements.
Only if Linux advocates and developers take a realistic look at their product offerings and their standing in the market.
The real problem with Linux is that the last 10% of any software is a real PITA. It's the kind of work that only gets done because you crawled out of bed on Monday so you don't get fired and are bored of staring at your monitor pretending to work. The kind where you reluctantly finish it because you're going on vacation next week and the only thing worse that actually finishing it is coming back to it afterwards.
I always have to laugh when I read news about EU suing Microsoft for bundling a browser or a media player with Windows, but fails to see the real issue - Microsoft's complete stronghold over hardware manufacturers.
For me the real issue is that Windows XP runs better on my netbook than linux.
Power: 7 watts lowest measured in XP, 10 watts lowest in Ubuntu 9.04. Noise: silent CPU fan in XP, noisy in Ubuntu Graphics: smooth scrolling in firefox in XP, molassas in Ubuntu (with or without compiz). Special Keys: all working in XP, only screen brightness in Ubuntu. Suspend: worked fine in XP, in Ubuntu the wifi module had to be unloaded and reloaded after a suspend or it would see networks but not transmit.
I finally got the power roughly equal by compiling an app not in any repository (with a broken make install) that underclocked and reduced the fan speed, and had a GUI for manually setting the keyboard commands (audio, wifi on/off, etc). But it was a PITA, and the graphics still weren't as snappy as in XP.
I use linux every day at work and I really wanted to use it on the netbook because I have XP already on a desktop for games. But the fact of the matter is that XP is better on the netbook than linux and for somebody that's never used linux before there's just no contest. To be 'fair', the vendor had pre-installed a couple utilities they wrote specifically for the hardware. But they could. How especially do they write the utils and support them for Ubuntu, and Fedora, and SuSE, and Debian, and...?
I have had countless Sun consultants for the best part of ten years telling me that Linux is unstable versus the 'rock solid' Solaris... When I challenge them for specifics they clam up tightly
I work a lot in VMs and linux is shit for stability when normal hardware expectations aren't met. Like the CPU frequency changing without Linux knowing about it, or having the clock instantly change by a couple months, or having disk IO be instant or take forever. Some specifics:
* A fedora core 7 snapshot that you can type a couple words into a terminal after reverting and then it slowly starts to lock up.
* Loaded a vmware image over an SMB network drive; linux guest crashed. Tried again and it worked ok.
* Frequently when I revert a fedora core image and click a close box, the box next to it is clicked. Or mouse clicks get doubled.
Lots of other instances of Linux bugging out like this. I don't know if Solaris has these problems, but I seriously doubt it. I think if you have actual hardware go bad you'll see a different side of Linux.
Common Lisp scores really well on benchmarks because they turn off all the safety protections that are most of the reason to use a dynamic language in the first place.
All scripting/dynamic languages are slow, and they always will be because nobody wants a really high level language that crashes.
I do a lot of my work in VMs and can tell you that the most important things, in order are:
1) RAM 2) RAM 3) Disk speed
You need a 64-bit CPU and you want it to be dual-core at least. But other than that it's basically a minor issue. All CPUs are 64-bit now except netbook Atom. VMs generally run say 10% slower in terms of CPU speed, if that, so just choose a CPU slightly beefier than what you'd need if you weren't using VMs.
Why is RAM so important? Because IO is a bottleneck normally and with VMs you are just creating more IOs so you want to avoid any swapping of any kind and you want everything in cache that can be. If you do any snapshots, with lots of RAM you can save/load to RAM; writes will be almost instant and reads *may* still be mostly in cache, if you're reverting a lot to the same snapshot.
Disk is the next most important thing for most of the same reasons. The slow part is always going to be IOs, and the ones you can't avoid doing by having huge ram need to be done asap. My advice would be to drop a couple hundred on an Intel SSD (or one of the better cheapo ones like OCZ Vertex for instance), make your VMs smallish so they fit on it, like 5-10gb, and if you need more space mount it. But read anandtech review first on SSDs, because some SSDs are worse than hdd! Furthermore, with snapshots there might be several actual IOs to find the block, since it might be in the main hdd file, or the per-snapshot changes files, etc. So again, you want blazing IO. Raid isn't worth doing, just get SSD. Plus, when you are done experimenting you can make use of SSD more than any other component you might buy to make the ultimate VM system.
Without the spiritual part of BSG it would have been just another space opera, probably fun to watch, to entertain, but certainly no to make you think.
But the spiritual part of it is what makes you NOT think. The episodes about torture of cylons, that made you think. The episodes about maintaining government, that made you think. The episodes about their despair and how they coped with it, that made you think. The questions of what you do with a person in charge forced to betray their people, that made you think. All the bad decisions people made because of their human emotions, that made you think.
But the episodes and stories about the gods and their plan... how does that make you think at all? You just listen to the story of it and that's it.
And for that matter, in the end, there's a monotheistic Christian-ish god, visions, angels, an afterlife, and a divine plan. Yeah, something to think about all right... basically the whole series turned out to be a really long episode of Touched by an Angel. No thanks.
Slashdot Mirror
am surprised at the amount. I figured it would be reduced to be more reasonable. My big problem with all this is the damages. $18,000 per song is 900 CD sales per song at $20 a CD.
Actually the point of statutory damages is precisely to be out of proportion to the actual damage, to prevent willful copyright infringement as a standard practice. If say you could only be fined twice the actual amount of damages, you could run a business infringing copyright if only half of the copyright owners didn't sue. And except for the extreme cases they wouldn't sue because it wouldn't be worth their time and money to do so.
For example if you made $5k in a year from selling illegal copies of Windows, it would cost far more than a $10k limit for Microsoft to sue over it. So that isn't the issue at all... 'ridiculously' high statutory damages are fine.
The issue is that penalty is completely out of proportion to the person's assets. There are two problems here 1) that it's based on number of copies instead of that copies were made and 2) it's based on a fixed amount of damages instead of based on assets.
A solution is simple... make the fine for wilful violation in addition to actual damages be 0.75% to 150% of the guilty party's assets instead of $750 to $150,000. Second, make the statutory damages be based on whether copyright infringement occurred instead of how many times. So in this case you'd have 24 songs @ $1 each or $24 plus. Then this person has say 90k in assets of their estate that's an extra $675 to $135,000 total depending on how bad the jury finds the violation was. This probably isn't possible within our system, but if you share 24 songs from different rights holders then they have to fight over that 0.75% to 150%.
Ok, that's pretty funny on both counts, but I often type in a URL into a search engine to make sure I go to the legit site and not accidentally to some "whitehouse.com" or "bankfoamerica.com" type of site.
Maybe I'm just some old timer and the anti-phishing in browsers is plenty fine... idk. But anyway, the point being that all the related searches and 'this is like' sidebars can actually help a lot. Another poster says goog does this also if you select the options to do it. So anyway I'm not trying to claim Bing invented anything (it's Microsoft...), just saying that as an the overall search engine it's on par with google if not better in some ways... definitely not '10 years behind' like the original poster said. I guess I missed the joke.
Bing gives us what Google already gave us 10 years ago. This is a major advance for Microsoft.
I think that's a little bit disingenuous. Try searching "reddit.com" of google and bing.
Google gives you a list of all results mentioning reddit.com, and a few common links into the site. That's it.
Bing gives you just the entry for reddit.com (probably what you want), and the common links into the site. There's a sidebar with related searches, "reddit nsfw", "reddit game", "twitter", etc. There's a sidebar that says similar to this site is digg, drudge report, huffington post, perez hilton. Judging by experience that's a really accurate summary of reddit.com. You can click 'show all' to see other pages that match "reddit.com".
Frankly I'm pretty impressed with bing, and I can see why google would be looking at it with a keen eye.
Also what irritates me even more than just bad results (which is understandable that search isn't going to always hone in on exactly what I'm looking for), is when the keyword isn't even in the results. Even if I put the "keyword" in quotes it does the same thing... links 'pointing to the page' contained that keyword. I think there is some type of operator:keyword to actually require the word, but who knows.
Code reviews are like testing... it's only worth doing if the quality is already high. A tester that ever has to file a bug like "crashes if run with zero args, or 5 or more args" is going to shoot somebody or just phone it in. Same with code reviews. If you have to explain to somebody why their use of a bubble sort is "not a good solution" then there is no point in doing one in the first place. It's just wasting people's time.
But on the other hand if you go to a code review and see some code implementing chmod octal and oug+-rwx,... syntax in 40 lines compiling to ~200 bytes of assembly you're going to say 'damn, that's cool' and it'll pump everybody up and they might learn a few things about binary operations and masks. You might catch some errors too, that's nice, but the main point of the review should be to learn about the code and to learn better programming. If you want to use them to catch errors, you should use testing and auditing of the code instead (if the presence of a bug is important enough to convene a meeting to look the for the mere possibility of it existing, then you need to really get serious about it).
I would assume every OS has an API to do this, but it might not be as fast as you want. Certainly on a modern Linux you can look at columns 6 and 7 in /proc/schedstat for each CPU and get a pretty good idea on CPU idleness.
That's not an "API", that's a hack workaround. Open a file, have kernel format the lines, read the lines, parse the lines, figure out how many CPUs to use, create a thread, use sched_* syscalls to assign it to a CPU, ... now CPU is busy again. Whoops. Even if you're using a thread pool it's so clumsy you might as well just assign it to N threads to begin with (where N is any arbitrary number). On top of that, it only works for one program at a time in the entire system (well best case it might be described by some kind of differential equation).
As far as I know, no OS has an API to do anything close to what I suggested in the post, a way to do fine-grained parallelism well. Maybe on a mainframe... Tera MTA could do this, but it had special hardware threads built into the CPU itself.
I've done some kernel work myself, and would try this, but it's not my job. And getting something like this past Ingo...
I can guarantee you that when you're at 35k feet and the speed sensor says you're going -1 kph a human pilot isn't going to aim the plane vertical and accelerate.
The big red override button sounds pretty good to me. If only they had air brakes.
I am not sure how you think that the splitting in itself would ensure immediate scheduling.
By definition. If the OS can't schedule it immediately on 4 cores, but can on 3 cores then it can split into 3 threads. No OS I know of can do this or has an API to do this.
Lo and behold, many OSes or low-level libraries also provide readily available thread pools!
Which you have to signal, then wait for the threads to be scheduled, then wait for the 'main thread' to be signaled that they completed. And you have no idea what else is going on so you don't know how many threads to use in the first place. ... which was kinda the point of my post in the first place.
Any program with a "for" or "while" loop in which the results of one iteration do not depend on the results of the previous iteration ... We just need the languages not to make coding this way too painful.
It's not the languages that are the problem, it's the operating systems. Check it:
1. Start 4 threads because OS said there are 4 cores
2. Each thread starts doing N/4 iterations of the loop
3. Each thread does some kind of synchronization to say it exited. It could be just a write, but there needs to be some way to wake up the original thread.
4. The original thread blocks until all the other threads have completed, then continues operating.
The best case is that there is lots of overhead creating a new thread, and the other threads have completed before the original one so that it doesn't have to wait.
The worst case is that the new threads don't get scheduled right away. Maybe flash is using 100% CPU on one core, so one thread might not even start running until one of the others completes. Anyway you end up with several threads not running and being scheduled at some later time. The for loop ends up taking several times longer even though it's running 'in parallel'.
Now what about a case like this:
1. ask os to split thread across multiple CPUs. This would be a guarantee that the threads would run immediately on the number of CPUs returned by the call.
2. run the loops on N cores (however many the OS returns).
3. each thread 'exits' when it is done. For the last thread to exit, the OS returns to it instead of exiting. Whichever thread finished last continues.
This doesn't suffer from the threads originally not starting right away and has no real synchronization except in the OS itself. It's awesome for smallish loops. It's also impossible to do with current operating systems.
The purpose of business is to make money. Not to be a patriotic cash funnel that supports governmental pet programs. Keep viewing corporations as ATM machines and they *will* relocate to more desirable locations because there are a lot of countries out there that see the benefits of all the jobs that large companies bring.
Business use a lot of government resources. The fire marshal gets called out to check that construction is following the fire codes, roads have to be built to their buildings and maintained and expanded. Water and sewer.
If you are using shared resources then you should be paying for them, in one way or another, and we have taxes to avoid the overhead of paying and accounting for every little detail by itself. Somebody has to pay to get things done, so if you are really against taxation on idealistic grounds you must be for companies paying for each and every service they use... including hard to quantify services like police and military protection.
No, a government job is not a "good" job, it is a drain on the tax base because it generates no wealth.
Generating wealth is irrelevant unless you can also protect it. If you get paid 10k and get robbed of it all then you might as well have just done nothing. Government jobs are these sorts of jobs that protect wealth or provide the support to enable others to generate wealth. You can't make a car with only an engine, and even if you try to maximize the engine by taking off "drains" like bumpers, the outer shell, upholstery, windshield, etc because theses don't 'generate' anything that doesn't make it a 'good' car.. that actually makes it a much worse car.
2. "gzip bytes," which they define as follows: "We started with the source-code markup you can see, removed comments, removed duplicate whitespace characters, and then applied minimum GZip compression." ... actually #2 seems to me like a very reasonable measure of expressiveness
I'm not so sure about that. For instance, their preprocessing effectively gives Python free blocks, by removing repeated spaces but not removing { and } from other languages. What they should do is:
1. remove all comments
That's it. You don't need to remove whitespace at all... gzip will compress 10 spaces to roughly the same as 1 space, if it's regular, but reducing 4, 8, 12 runs of spaces to 1 will affect the compression. You don't need to rename variables, because AReallyLongVariable used a dozen times will be amortized away if it is used often... it won't affect the gzip size much more than calling it 'a1'. They should also use maximum compression (that's the point of compressing it in the first place, to reduce it to the essential structure)... they should use 7zip or rar. So IMO their preprocessing of the source is creating a misleading metric.
Wolfram Alpha is for when you want to know how long it will take for two orbits around the galaxy. Also, how long it takes the solar system to orbit the galaxy is not what I'd call 'useful' information...
Nobody wants to sell their pride and joy, their little tech baby. That's not the question. The question is should they sell it.
Ultimately there are just two factors to consider:
1. How much are you really going to get paid?
2. Will it still be your pride and joy?
Both are completely situational. Do you get paid up front in addition to maybe being an employee? How much do you expect to lose by selling it? Does the company want to develop the product, or shut it down? Megacorp might fire you right away, so make sure to get a guarenteed payment of some kind that is substantial. You might be able to make more money off your code yourselves, or you might fail. Nobody here can tell give you any advice, because you gave us no information.
But I would tend to say that guaranteed money is better than potential money. Also remember human nature; people that have their ships come in and get lump sums as a rule end up blowing it. You don't know what kind of taxes you get when you get a 100k lump sum (or whatever), or how fast it's going to burn. Don't go out and blow it right away on cars and fancy houses, or decide to become a day trader!
fmpeg decodes nearly everything, I don't think I've used a binary codec in quite a long time... MPEG-2, MPEG-4, H.264, .flv, wmv9, even realvideo files, ffmpeg does it natively.
Just google around a bit, for instance a post from Oct 2008: "libavcodec has some ARMv6 optimized stuff, but nothing for ARMv7 or TMS320C64x. Looks like a good project for anyone who is familiar with those"
Look at the libavcoded source... there's way more ASM for x86 and it looks like a lot more work has been done on x86 optimizations, which is to be expected. And there are still other codecs not supported at all.
it's a revelation -- you realize quick, there's not that much on a linux desktop that relies on x86.
Linux has basically zero desktop market share, so nobody really cares if their linux programs work. They care if their Windows programs work.
I've heard there's a flash for arm so that's not missing.
"I've heard" gets you +informative? Gimme a break. Adobe doesn't have a download link for flash for ARM, and linux flash copycats are buggy(er) and slow. "Not missing" and "just as good" are completely different.
And yet I hear people talking as though ARM-based netbooks will be on the market within a year or something. What am I missing here? Is it all vaporware?
It's ridiculous. People say ARM is so much more power efficient, then forget that they have to emulate x86 because some video codec isn't written for ARM. A big reason people bought linux netbooks was to put bootleg Windows on them, to run Windows apps. Windows has like 90% of the netbook market now. ARM would have to be ten times better architecture than x86 to even make a dent in netbooks, and it's barely even 1.12 times better.
Hell, as phones get even more powerful we'll probably see the opposite... x86 invading phones. After all, how much cooler would an iPhone be if you could run crysis on it?
I designed one of these about a decade ago and did some prototyping. Since I don't seem to have the time to realize it, here are a few extra features that could be added (if i2p does not already include these).
Here's a tip for anybody thinking of implementing something like i2p, tor, freenet, etc: if the user has to do anything and if it impacts performance it's not worth doing.
What's needed is something simple and pervasive:
1) compatible with regular TCP
2) optional so it is only in effect when both the source and destination support it
3) 'weak' so that there is not much performance impact, so there is no reason to disable it
For sake of example, half of an xtea key can be sent by the SYN using the TCP options field and the other half provided by the server in the ACK. If the server doesn't return its half then no encryption takes place. The key is permuted by the data sent/received.
This simple scheme provides that anybody examining the TCP steam must have seen the first packets and must have followed the entire conversation, decrypting it along the way. There is no extra step involved that could introduce delays, and the state and CPU time is small enough to be not important to the end user or server. However, for anybody to wholesale monitor traffic, for whatever reason, it means they need a lot of expensive hardware. What ends up as 1% of your CPU time to decrypt a torrent requires a warehouse of servers to decrypt everybody's torrents.
A system like this has a huge advantage over tor, freenet, etc, in that everyday normal people can have it enabled by default, especially for open-source linux, *bsd distros. The actual anonymizing networks are worthless because only those with something to hide use them, or people who are hard-core idealists (which probably also gets you on some kind of 'watch list'). Ironically, this kind of system will raise the overall cost of monitoring to a point where tor, freenet, i2p, etc become viable.
In firefox I can go full screen, but still keep the URL bar, so I know what site I'm on. I can get a menu bar by pressing Alt once. I can put the NoScript button on the URL bar. I can even put a button to toggle fullscreen, since it's often easier to trackpad to a button than find F11 on a small non-backlit keyboard. Or firefox can drop down this UI when the cursor is at the top.
Chrome can do none of these things, even in the current beta version. The UI in many ways is restrictive and lacks many niceties. You can still like it and prefer it, that's fine, but let's be honest. Perhaps I should have said "maximize space while still being usable", but some common sense is in order here.
Chromifox makes firefox look a lot like Chrome. Chrome is a nice toy, but it's UI is pretty lacking when you want to do something like maximize screen space on a 1024x600 screen.
The other thing about linux on a netbook is: firefox 3 runs javascript like a dog, and that really shows up when you don't have a beefy machine.
It's not javascript, or not just javascript. I've seen xulrunner apps that are magnitudes faster in Windows than in Linux. Apparently the rendering in Linux is not clipping correctly, so it spends a lot of extra time drawing hidden elements.
Only if Linux advocates and developers take a realistic look at their product offerings and their standing in the market.
The real problem with Linux is that the last 10% of any software is a real PITA. It's the kind of work that only gets done because you crawled out of bed on Monday so you don't get fired and are bored of staring at your monitor pretending to work. The kind where you reluctantly finish it because you're going on vacation next week and the only thing worse that actually finishing it is coming back to it afterwards.
I always have to laugh when I read news about EU suing Microsoft for bundling a browser or a media player with Windows, but fails to see the real issue - Microsoft's complete stronghold over hardware manufacturers.
For me the real issue is that Windows XP runs better on my netbook than linux.
Power: 7 watts lowest measured in XP, 10 watts lowest in Ubuntu 9.04.
Noise: silent CPU fan in XP, noisy in Ubuntu
Graphics: smooth scrolling in firefox in XP, molassas in Ubuntu (with or without compiz).
Special Keys: all working in XP, only screen brightness in Ubuntu.
Suspend: worked fine in XP, in Ubuntu the wifi module had to be unloaded and reloaded after a suspend or it would see networks but not transmit.
I finally got the power roughly equal by compiling an app not in any repository (with a broken make install) that underclocked and reduced the fan speed, and had a GUI for manually setting the keyboard commands (audio, wifi on/off, etc). But it was a PITA, and the graphics still weren't as snappy as in XP.
I use linux every day at work and I really wanted to use it on the netbook because I have XP already on a desktop for games. But the fact of the matter is that XP is better on the netbook than linux and for somebody that's never used linux before there's just no contest. To be 'fair', the vendor had pre-installed a couple utilities they wrote specifically for the hardware. But they could. How especially do they write the utils and support them for Ubuntu, and Fedora, and SuSE, and Debian, and ...?
I have had countless Sun consultants for the best part of ten years telling me that Linux is unstable versus the 'rock solid' Solaris ... When I challenge them for specifics they clam up tightly
I work a lot in VMs and linux is shit for stability when normal hardware expectations aren't met. Like the CPU frequency changing without Linux knowing about it, or having the clock instantly change by a couple months, or having disk IO be instant or take forever. Some specifics:
* A fedora core 7 snapshot that you can type a couple words into a terminal after reverting and then it slowly starts to lock up.
* Loaded a vmware image over an SMB network drive; linux guest crashed. Tried again and it worked ok.
* Frequently when I revert a fedora core image and click a close box, the box next to it is clicked. Or mouse clicks get doubled.
Lots of other instances of Linux bugging out like this. I don't know if Solaris has these problems, but I seriously doubt it. I think if you have actual hardware go bad you'll see a different side of Linux.
Common Lisp scores really well on benchmarks because they turn off all the safety protections that are most of the reason to use a dynamic language in the first place.
All scripting/dynamic languages are slow, and they always will be because nobody wants a really high level language that crashes.
The most important bits are your CPU and RAM.
I do a lot of my work in VMs and can tell you that the most important things, in order are:
1) RAM
2) RAM
3) Disk speed
You need a 64-bit CPU and you want it to be dual-core at least. But other than that it's basically a minor issue. All CPUs are 64-bit now except netbook Atom. VMs generally run say 10% slower in terms of CPU speed, if that, so just choose a CPU slightly beefier than what you'd need if you weren't using VMs.
Why is RAM so important? Because IO is a bottleneck normally and with VMs you are just creating more IOs so you want to avoid any swapping of any kind and you want everything in cache that can be. If you do any snapshots, with lots of RAM you can save/load to RAM; writes will be almost instant and reads *may* still be mostly in cache, if you're reverting a lot to the same snapshot.
Disk is the next most important thing for most of the same reasons. The slow part is always going to be IOs, and the ones you can't avoid doing by having huge ram need to be done asap. My advice would be to drop a couple hundred on an Intel SSD (or one of the better cheapo ones like OCZ Vertex for instance), make your VMs smallish so they fit on it, like 5-10gb, and if you need more space mount it. But read anandtech review first on SSDs, because some SSDs are worse than hdd! Furthermore, with snapshots there might be several actual IOs to find the block, since it might be in the main hdd file, or the per-snapshot changes files, etc. So again, you want blazing IO. Raid isn't worth doing, just get SSD. Plus, when you are done experimenting you can make use of SSD more than any other component you might buy to make the ultimate VM system.
Without the spiritual part of BSG it would have been just another space opera, probably fun to watch, to entertain, but certainly no to make you think.
But the spiritual part of it is what makes you NOT think. The episodes about torture of cylons, that made you think. The episodes about maintaining government, that made you think. The episodes about their despair and how they coped with it, that made you think. The questions of what you do with a person in charge forced to betray their people, that made you think. All the bad decisions people made because of their human emotions, that made you think.
But the episodes and stories about the gods and their plan... how does that make you think at all? You just listen to the story of it and that's it.
And for that matter, in the end, there's a monotheistic Christian-ish god, visions, angels, an afterlife, and a divine plan. Yeah, something to think about all right... basically the whole series turned out to be a really long episode of Touched by an Angel. No thanks.