While what you say is true, that Steam comes to Linux is infinitely better than it staying on Windows, also for people's freedom.
15 years ago I heard similar arguments that ports of free software to DOS/Windows should be discouraged because of the intermixing. Guess what? That intermixing gave me and a whole bunch of other people a hint at what free software really means and eventually brought us over.
So get off your high horse, mr. Anonymous Coward, and let people cheer.
While we're at John Carmack, he also said that he found the Intel open source drivers really interesting to look at, and recommended it to anyone else writing a graphics engine. He then proceeded to say that if he could clone himself so he would have more time, he would love to work on optimizing them.
So it's not just the Valve developers who see the benefit of open source drivers.
I hope they will at least break backwards compatibility in order to justify the change.
They did that in a clever way with 3.0. See, many scripts were relying on the 2.6.x numbering scheme, or at least a triple scheme of x.y.z, so just by changing the version number they broke a lot of stuff. Clever!
If UnionFS is not the answer, and OP doesn't want something more complicated, it honestly sounds like the simplest solution is to just symlink everything into another directory once it's up running.
You have to do something about failures, but if you have that in hand, making the symlinks is really easy, you can do it with three lines of bash (for each path in find -type f, create symlink). If you need a bit more control, and it's not easily doable with grep, I'd write a small Python program, you can google up the necessary directory recursion code easily enough.
This would probably be much easier to understand than some auto-magic FS stuff. It means you will have to read the entire directory structure, but if you going to do that anyway, pulling it into the disk cache at boot is perhaps not a bad idea? If you start the script once for each disk, you can do the reading in parallel.
If that's fine, all you need is a distribution with the remote mount stuff set up. It's not too hard to do on Debian, but you will have to edit a configuration file or two.
Good point, but for once they're actually just offering VPSes it seems, so you can just rsync your stuff out. Of course, if you have some custom scripting set up for starting instances, you'll have to rewrite that since the APIs will be different. But that's probably only a small hassle.
Steep if you need it 24/7. These kind of offerings are mostly a good deal if you either use them for temporary computations, so you spin up a bunch of machines, do your thing and then close them down again (so you may end up paying for 50 machines for 5 hours which is definitely cheap), or if you are in a web startup that has venture funding with lots of cash to burn and potential day-to-day scaling problems.
Otherwise there are better options, for instance Hetzner that'll offer better hardware for ~50 USD/month (42 EUR ex VAT for a non-virtualized i7 with 16 GB RAM and 3 TB RAID1).
I clearly recall one week my three kids caught the same flu and I spent $900 out of pocket for 3 doctors visits.
Just curious, but why do you go to the doctor for a flu? They can't really do anything about it anyway? Worried about the temperature getting too high?
If you move to Denmark, doctor and hospital visits will be 100% free no matter what, dental free only up to 18 years (but I think they're looking to change that), if you got something serious I think you can get some money for the medicine, but otherwise you'll have to pay for that yourself. Same with glasses, birth control pills and that sort of thing. If you get old, things change so you can actually some coverage of glasses and hearing aids.
I don't know how things work for foreigners, but I imagine as soon as you get a permit to stay here, you're covered in the same way as Danish citizens.
Damn, I forgot to tell my anecdote - my girlfriend's original Logitech finally died one day, but at that time she had inherited a computer from her brother, with the same goddamn keyboard.
I managed to kill that keyboard by accidentally spilling half a glass of milk over it last week, despite washing and taking it apart for drying. So eventually she had to accept getting a new one.
My girlfriend had a Logitech keyboard. Forgot the model name, probably middle-of-the-road. Anyway, the keypress action was horrible. Sometimes when you didn't hit a key spot on, it would go down alright but just not register the press.
Hence, I'm staying away from Logitech keyboards.
At work, we bought Microsoft Comfort Curve 2000 when we started (15-20 USD). Horrible build quality, especially space bar, but really nice to type on. Ctrl is in a much better spot for Emacs'ing. They don't seem to be good for more than 3-6 years of everyday typing, though. I recently bought a Microsoft Comfort Curve 3000 (also 15-20 USD), and so far am liking it despite some annoying innovations (like insert being put away).
Why do you read (or rather listen) to books you don't like?
I think the only time I've read a book to feel cool was The Unbearable Lightness of Being by Milan Kundera and that was actually interesting (and also a relatively short book).
My girlfriend sometimes borrows "in" books and usually never gets through them - e.g. I think she's been trying to get through The Lord of the Ring by Tolkien three times. Well, I did actually get through that 10 years ago, but only by speed-reading the whole life-is-so-hard-we-got-to-get-through-this-bog thing with Frodo. Anyway, I don't understand why she does it. That something is a classic just means that some people at some point really liked it, not that you will. If you don't like it, drop it.
Have you any actual experience with a networked server? The actual protocol is easy enough to do. Sending data from one party to another and parse it is easy.
The real problem is that you go from a situation where all components can talk to each other quickly to a situation where everything you do must be carefully analysed because each time you cross the boundary between local and remote you take a possibly big hit from latency and possibly also bandwidth if it's a lot of data.
I have a Master's degree in distributed systems, and the mere term "network-transparent" is probably a fallacy. In face of the huge performance implications of a network, the local-remote split most likely has to be explicit in some way, just like it is with web. Unless you don't crappy performance.
If you storing it uncompressed, you are most likely wasting space. Even very slight compression will be invisible in practice* and can cut down the file sizes dramatically. You can easily get down to 25-50% with no loss in visual quality. If you don't believe me, do try it.
(*) - note that what matters to people looking at the photos is practice, not theory. What matters is also that any manipulation of them is 3-4 faster. I am not talking about blocky overcompression like we had in the nineties (and still see today in crappily compressed video streams). I'm a talking about compression that you cannot see, like high-rate lossy audio compression that is indistinguishable to the originals in double-blind tests by audio experts.
Interesting calculation, but I think you're forgetting that you will be looking through the photos anyway. Thus deleting will only add a fraction of overhead, and can actually improve browsing time if you ever need to look at it again since you don't have to look at all the total crap.
If the buyer is willing to buy at 3 and the seller is willing to sell at 1, they should meet at 2.
And this happens exactly how? They start negotiating over a cigar and a bottle of rum? Have you ever actually tried this?
With a stock exchange, what happens here is that either the buyer comes first in which case the seller sees his offer and sells for 3, or the seller comes first so the buyer sees his offer and buys for 1 (subtracted trading fees of course). They will never meet at 2. So that's why one of those poor sods will always come out with -1 while the other gets +1 compared to the average.
If you add a market maker, you might have the market maker buying at 1.99 and selling at 2.01. In this case whoever comes first, unless they accept a worse deal than the market maker is offering, both will come out with -0.01 compared to the average.
Now realize that the guy with the advantage here will usually be a big bank or investment company who has a much better idea of what happening in the market, and I would gladly take the -0.01 hit any day compared to the risk of the -1.
Why is the exchange itself not matching up buyers with sellers? why do we need these third party traders doing it?
While I am no expert on how exactly various algorithms earn money, it's my impression that some algorithms can make money by smoothing out differences between different markets (stock exchanges are much worse connected than you'd think in today's world, e.g. from my local Danish bank I can't invest online in American stock exchanges, only the Danish one), but market making generally involves risks since you buy something, hold on to it and hope to sell it for more when a new buyer shows up later. If the market falls, you're in trouble. That's why you don't want the stock exchange to do it.
If you think a market maker is a useless middle man, then think of the case where you want to buy something, but nobody wants to sell. Either you pay a big premium, or you rejoice because a market maker is willing to sell at a small premium because they bought up some stock earlier at a small discount. The loser here is the guy who could have sold to you at a big premium, taking advantage of the illiquidity. Thus you benefited and so did the guy who sold the market maker the stock because he didn't have to either wait for his sale or sell at a huge discount. And if anyone was willing to pay a better price than the market maker at the time, he could have sold to him instead.
You certainly want algorithms to do market marking - there's just very little point in having a human monitor the value and putting in orders, it just raises the spread necessary to pay for that guy. Instead various companies can put up algorithms, compete and lower the spread even further.
I think the high-frequency aspect is silly too, but I haven't heard any credible cases yet where someone who is not taking advantage of another party actually loses money. And I think the millisecond aspect of it is more about beating the other HFT algorithms to the punch rather than stealing deals from humans.
That all sounds very good, until you realize that the HFT is just playing the part of a middleman, adding, well, no value to the exchange. Without the HFT the buyer and the seller would just talk to each other, negotiate a price, and the seller would get more for his stock while the buyer paid less.
Could you please explain how that happens? You are not describing the stock exchanges I have used. In there, you either post an offer or is being matched with an existing offer. A HFT will only have a chance if can offer me a better price than everyone else.
As I understand things, market making generally lowers the spread, which is bad if you're trying to use an illiquid market to gain a price more advantageous than what would be fair was the market more liquid.
A 1GW-rated nuclear plant will produce 1GW 24/7/365, rain or shine.
No, it will not. Maybe an imaginary one, but not a real plant.
A real city, complete with homes and businesses and some industry. If they can't do that, then that power source will never be good for more than a few percent of national power usage.
In Denmark, we passed the 15% mark some years ago. But I guess we live in a fantasy world.
I think your position is silly - the future of renewable power sources isn't in being disconnected from the grid, au contraire. From what I can find, people who've actually done some real calculations on this say that it's possible to rely 100% on renewable sources without excessive costs. And why wouldn't it be, we have the tech.
Slashdot Mirror
While what you say is true, that Steam comes to Linux is infinitely better than it staying on Windows, also for people's freedom.
15 years ago I heard similar arguments that ports of free software to DOS/Windows should be discouraged because of the intermixing. Guess what? That intermixing gave me and a whole bunch of other people a hint at what free software really means and eventually brought us over.
So get off your high horse, mr. Anonymous Coward, and let people cheer.
While we're at John Carmack, he also said that he found the Intel open source drivers really interesting to look at, and recommended it to anyone else writing a graphics engine. He then proceeded to say that if he could clone himself so he would have more time, he would love to work on optimizing them.
So it's not just the Valve developers who see the benefit of open source drivers.
While that may be true today, Google is probably building this stuff so that when the world is ready, Google will be there for them.
Actually, if they could just tap your head to fire at it, you might be in for some trouble. :)
I hope they will at least break backwards compatibility in order to justify the change.
They did that in a clever way with 3.0. See, many scripts were relying on the 2.6.x numbering scheme, or at least a triple scheme of x.y.z, so just by changing the version number they broke a lot of stuff. Clever!
Great post, thanks!
If UnionFS is not the answer, and OP doesn't want something more complicated, it honestly sounds like the simplest solution is to just symlink everything into another directory once it's up running.
You have to do something about failures, but if you have that in hand, making the symlinks is really easy, you can do it with three lines of bash (for each path in find -type f, create symlink). If you need a bit more control, and it's not easily doable with grep, I'd write a small Python program, you can google up the necessary directory recursion code easily enough.
This would probably be much easier to understand than some auto-magic FS stuff. It means you will have to read the entire directory structure, but if you going to do that anyway, pulling it into the disk cache at boot is perhaps not a bad idea? If you start the script once for each disk, you can do the reading in parallel.
If that's fine, all you need is a distribution with the remote mount stuff set up. It's not too hard to do on Debian, but you will have to edit a configuration file or two.
Good point, but for once they're actually just offering VPSes it seems, so you can just rsync your stuff out. Of course, if you have some custom scripting set up for starting instances, you'll have to rewrite that since the APIs will be different. But that's probably only a small hassle.
Steep if you need it 24/7. These kind of offerings are mostly a good deal if you either use them for temporary computations, so you spin up a bunch of machines, do your thing and then close them down again (so you may end up paying for 50 machines for 5 hours which is definitely cheap), or if you are in a web startup that has venture funding with lots of cash to burn and potential day-to-day scaling problems.
Otherwise there are better options, for instance Hetzner that'll offer better hardware for ~50 USD/month (42 EUR ex VAT for a non-virtualized i7 with 16 GB RAM and 3 TB RAID1).
I clearly recall one week my three kids caught the same flu and I spent $900 out of pocket for 3 doctors visits.
Just curious, but why do you go to the doctor for a flu? They can't really do anything about it anyway? Worried about the temperature getting too high?
If you move to Denmark, doctor and hospital visits will be 100% free no matter what, dental free only up to 18 years (but I think they're looking to change that), if you got something serious I think you can get some money for the medicine, but otherwise you'll have to pay for that yourself. Same with glasses, birth control pills and that sort of thing. If you get old, things change so you can actually some coverage of glasses and hearing aids.
I don't know how things work for foreigners, but I imagine as soon as you get a permit to stay here, you're covered in the same way as Danish citizens.
Aw, come on, mod parent up, it's funny!
(Yes, I enjoyed model M too. My surroundings probably didn't.)
Damn, I forgot to tell my anecdote - my girlfriend's original Logitech finally died one day, but at that time she had inherited a computer from her brother, with the same goddamn keyboard.
I managed to kill that keyboard by accidentally spilling half a glass of milk over it last week, despite washing and taking it apart for drying. So eventually she had to accept getting a new one.
My girlfriend had a Logitech keyboard. Forgot the model name, probably middle-of-the-road. Anyway, the keypress action was horrible. Sometimes when you didn't hit a key spot on, it would go down alright but just not register the press.
Hence, I'm staying away from Logitech keyboards.
At work, we bought Microsoft Comfort Curve 2000 when we started (15-20 USD). Horrible build quality, especially space bar, but really nice to type on. Ctrl is in a much better spot for Emacs'ing. They don't seem to be good for more than 3-6 years of everyday typing, though. I recently bought a Microsoft Comfort Curve 3000 (also 15-20 USD), and so far am liking it despite some annoying innovations (like insert being put away).
Agreed. Thanks!
Why do you read (or rather listen) to books you don't like?
I think the only time I've read a book to feel cool was The Unbearable Lightness of Being by Milan Kundera and that was actually interesting (and also a relatively short book).
My girlfriend sometimes borrows "in" books and usually never gets through them - e.g. I think she's been trying to get through The Lord of the Ring by Tolkien three times. Well, I did actually get through that 10 years ago, but only by speed-reading the whole life-is-so-hard-we-got-to-get-through-this-bog thing with Frodo. Anyway, I don't understand why she does it. That something is a classic just means that some people at some point really liked it, not that you will. If you don't like it, drop it.
But that's just my humble opinion. :)
Enjoy!
I should have backed this.
you could play moonlight sonata on an iphone with no musical training. it's not the most complex of tracks.
You could, but it would sound so much better with a professional pianist with the right equipment. :)
Have you any actual experience with a networked server? The actual protocol is easy enough to do. Sending data from one party to another and parse it is easy.
The real problem is that you go from a situation where all components can talk to each other quickly to a situation where everything you do must be carefully analysed because each time you cross the boundary between local and remote you take a possibly big hit from latency and possibly also bandwidth if it's a lot of data.
I have a Master's degree in distributed systems, and the mere term "network-transparent" is probably a fallacy. In face of the huge performance implications of a network, the local-remote split most likely has to be explicit in some way, just like it is with web. Unless you don't crappy performance.
If you storing it uncompressed, you are most likely wasting space. Even very slight compression will be invisible in practice* and can cut down the file sizes dramatically. You can easily get down to 25-50% with no loss in visual quality. If you don't believe me, do try it.
(*) - note that what matters to people looking at the photos is practice, not theory. What matters is also that any manipulation of them is 3-4 faster. I am not talking about blocky overcompression like we had in the nineties (and still see today in crappily compressed video streams). I'm a talking about compression that you cannot see, like high-rate lossy audio compression that is indistinguishable to the originals in double-blind tests by audio experts.
Interesting calculation, but I think you're forgetting that you will be looking through the photos anyway. Thus deleting will only add a fraction of overhead, and can actually improve browsing time if you ever need to look at it again since you don't have to look at all the total crap.
Perhaps because you're afraid that something really bad happens that changes your assumptions while you look the other way?
If the buyer is willing to buy at 3 and the seller is willing to sell at 1, they should meet at 2.
And this happens exactly how? They start negotiating over a cigar and a bottle of rum? Have you ever actually tried this?
With a stock exchange, what happens here is that either the buyer comes first in which case the seller sees his offer and sells for 3, or the seller comes first so the buyer sees his offer and buys for 1 (subtracted trading fees of course). They will never meet at 2. So that's why one of those poor sods will always come out with -1 while the other gets +1 compared to the average.
If you add a market maker, you might have the market maker buying at 1.99 and selling at 2.01. In this case whoever comes first, unless they accept a worse deal than the market maker is offering, both will come out with -0.01 compared to the average.
Now realize that the guy with the advantage here will usually be a big bank or investment company who has a much better idea of what happening in the market, and I would gladly take the -0.01 hit any day compared to the risk of the -1.
Why is the exchange itself not matching up buyers with sellers? why do we need these third party traders doing it?
While I am no expert on how exactly various algorithms earn money, it's my impression that some algorithms can make money by smoothing out differences between different markets (stock exchanges are much worse connected than you'd think in today's world, e.g. from my local Danish bank I can't invest online in American stock exchanges, only the Danish one), but market making generally involves risks since you buy something, hold on to it and hope to sell it for more when a new buyer shows up later. If the market falls, you're in trouble. That's why you don't want the stock exchange to do it.
If you think a market maker is a useless middle man, then think of the case where you want to buy something, but nobody wants to sell. Either you pay a big premium, or you rejoice because a market maker is willing to sell at a small premium because they bought up some stock earlier at a small discount. The loser here is the guy who could have sold to you at a big premium, taking advantage of the illiquidity. Thus you benefited and so did the guy who sold the market maker the stock because he didn't have to either wait for his sale or sell at a huge discount. And if anyone was willing to pay a better price than the market maker at the time, he could have sold to him instead.
You certainly want algorithms to do market marking - there's just very little point in having a human monitor the value and putting in orders, it just raises the spread necessary to pay for that guy. Instead various companies can put up algorithms, compete and lower the spread even further.
I think the high-frequency aspect is silly too, but I haven't heard any credible cases yet where someone who is not taking advantage of another party actually loses money. And I think the millisecond aspect of it is more about beating the other HFT algorithms to the punch rather than stealing deals from humans.
That all sounds very good, until you realize that the HFT is just playing the part of a middleman, adding, well, no value to the exchange. Without the HFT the buyer and the seller would just talk to each other, negotiate a price, and the seller would get more for his stock while the buyer paid less.
Could you please explain how that happens? You are not describing the stock exchanges I have used. In there, you either post an offer or is being matched with an existing offer. A HFT will only have a chance if can offer me a better price than everyone else.
As I understand things, market making generally lowers the spread, which is bad if you're trying to use an illiquid market to gain a price more advantageous than what would be fair was the market more liquid.
A 1GW-rated nuclear plant will produce 1GW 24/7/365, rain or shine.
No, it will not. Maybe an imaginary one, but not a real plant.
A real city, complete with homes and businesses and some industry. If they can't do that, then that power source will never be good for more than a few percent of national power usage.
In Denmark, we passed the 15% mark some years ago. But I guess we live in a fantasy world.
I think your position is silly - the future of renewable power sources isn't in being disconnected from the grid, au contraire. From what I can find, people who've actually done some real calculations on this say that it's possible to rely 100% on renewable sources without excessive costs. And why wouldn't it be, we have the tech.