Monty seemed to be saying that not having to consider crap like OS buffering would make implementing an ACID database a lot easier...
But, they don't even know the OS they are programming to, so either you use O_SYNC when on posix, or you can't predict anything about VFS buffering. And even when using O_SYNC, you are in no way guaranted (when using pure IDE IIRC) in what order your commands will hit the disk, and when (on-disk cache).
So, I also have no idea w.t.f. the physical nature of the storage medium should have to do with ACID.
When you turn on this gadget, add your mailing list addresses to your white list. If you suddenly stop getting a list, go find out if they changed their sending address and add it to your white list.
Only to find out that the mailing list admin was fed up with umpteen "challenges" when he changed the sending adress and banned you for life.
Not to mention one of the biggest problems: Every spam message sent will consume the bandwidth it always has consumed, but will now trigger the C/R system to send a message back. So you have twice the email traffic. And have you ever been the victim of a spammer that used your email address as the From/Return-Path and you received all the bounces? Now imagine a spammer doing this and not only receiving all the bounces but also all the C/R requests.
Excellent idea, instant DOS attack:
From: support@microsoft.com To: everyone@earthlink.com Subject: GET A BIGGER (whatever) NOW!!!!!!
[...]
The funny thing is, that a system like this might _drive_ spammers to use From: adresses which the deem more likely to be whitelisted (esp. since the possibility of whitelisting complete domains seems to be a nice feature at first).
Now one spam message creates a reply which has 100 fold the size of the average spam message, and, since the mail is forged anyway, goes nowhere.
Worse, if spammers forge valid adresses, one poor sob get's 5 Gigs of useless pictures of "validation emails" in his inbox instead of 1000 hatemails from lusers accusing him of spamming.
At least they should send pictures of naked supermodels with the confirmation secrets tatood on their butts.
Why would any terrorist worry about getting boxcutters past airport security now when they could dump an assload of ricin into a big city's reservoir and watch hundreds of thousands of people croak?
Erm, obviously they decided to do the plane thing, although there were _no_ homeland security stuff and _no_ cameras installed at that time.
People, face it, there are a million things dedicated evil people can do and _nothing_ can really prevent it. America is _not_ at war (at least there's no war at its home turf - ask any nation who really was hit by war), and all this stuff does is to cause mass paranoia without gaining anything.
Oh wait, there are nice companies gaining from paranoia, but that's it.
OTOH, there's an interesting short story in one of Hofstaedter's (spelling right? to lazy to look up) books where one guy build mechanical "bugs" which could express very simple "emotions" - i.e. making some pet like noises, making noises of fear, crawinling around etc. - and he asks a visitor to smash one bug with a hammer. The story describes how this man wasn't emotionally able to do that, because he developed feelings for this bug. (IIRC the real story is somewhat more involved, but you get the idea).
I bet if you decorated an "intelligent" AI with some emotional dressing, you could significantly lower the barrier to accept it as "intelligent".
Shows how deeply involved the human perception of not only intelligence, but life in general is.
But doesn't that example illustrate rather a human "deficiency" than one from a machine? I mean the deficiency that there is far more "computation" going on in our brain than we consciously notice. I don't think that this "right feeling" was the result of something supernatural going on.
To add another example. Richard Feynman writes in one of his books about how he listened to a conversation between to students of mathematics(at MIT or so) where one tried to explain some mathematical concept to the other. Feynman described that after a long while and much intense explaining, finally the other student "got it", and said something along the lines of "Oh! YES, THAT'S TRIVIAL!" Feynman goes on to make fun of mathematicians by proposing that mathematicians only understand trivial problems, because anything they have already understood is declared trivial by them.
This is a bit extreme, but it decribes exactly the notion some AI critics seem to have when judging AI advances.
1) Make a computer with true free will. Let's see AI do something it wasn't originally designed to do because it wants to.
The problem is that philosphers/psycologists/neurologists/genome scientists are debating since centuries if and in what magnitude this sentence is true when you replace computer and AI with man.
Exactly. And since philophers and other scientists are not dumb, they even go to the root of the problem in advance. Searle's chinese room theory is a prime example for this IMO, you could use this reasoning to rip apart _anything_ which will come out of AI science. Note that I don't think that AI today is able to resemble anything which will really master a turing test, but even if it did, the chinese room argument will offer a way to argue that the computer "doesn't really understand what he is talking." I for one think that this argument is silly, because it relies on our inability (today) to deconstruct the brain in the same way we could with a software. If we could, we'd find out that the different "parts" of our brain don't really "understand" what they are doing - in the same way. Oh, and cars aren't really able to go forward, because an engine is not able to move on itself.
There is at least on userspace daemon which reach the same performance/scalability on linux (commercial, I don't remember its name, the creator posted on linux-kernel about it). The point is that when implementing tux, they made a lot of general improvements in the kernel. These are the reason for tux' performance, not the fact that it is in kernel space. Therefore the spec numbers of tux give far more substance to the scalability debate than 5 years old hearsay.
FreeBSD 5.0 seems superior to Linux in the role of webserver when it comes to scaling,...
This myth has to die or someone has to prove it, with recent versions of linux and bsd. There have been so many advances with linux (for instance after the mindcraft incident etc.) that I doubt it's true. The current advances in linux can clearly be seen over at spec.org, linux with the tux webserver scales nearly linearly to 8 procs.
I didn't want to write too much. First, a player like Microsoft certainly can buy enough marketing power, and second, if the internet gets a more important distribution channel for a good, internet specific marketing will get more important for that good. A lot of music marketing consists of paying shops for putting albums in prominent position, shipping cds in as much shops you can, paying shops for displaying your cd in their advertisements etc, which is all related to the classical distribution methods.
Oh, and that thing about taking risks... It's not that artists don't want to take the risk, they just can't afford it. If they had the money (or could get a credit), I'm sure most wanted to take the risk under the same circumstances (e.g. marketing power), because for all I read, making a contract with label is mainly an very expensive loan.
Wrong. The hardware and support costs can be amortized on a huge scale. How many hosting centers is this? How many distribution centers are CDs shipped from? Do the math, its not even close. Record companies are going to make a killing on cost davings alone if this pans out.
I think and you are wrong, and that you are wrong for the same reason record companies are trying to get draconian with copy protection.
If this takes off and record companies enter the game in a big way, it will take off big - very big. So big that it has the potential to badly damage the classic distribution chains. Music is the optimal good for distribution over the net in the state it has today (average bandwidth for the end user).
Fast dsl/cable connections make the act of purchasing and downloading music in a compressed format unpremeditated buying.
After online sales getting a significant share of the total revenue, there's suddenly a very low barrier of entry for anyone for this business.
Why do musicians sign their soul to big music companies? Because they are the only one offering the the things they need (or believe they need in case of the first):
- marketing power - logistics (they can make an album appear in every shop on the planet)
It's clear the internet solves the logistics, and this is IMO the biggest hinderance for newcomers. It also could raise the absolute number of sales (unpremeditated buying etc.).
But also completely new competitors could emerge, or artists might consider handling their own sales, which all will eventually drive down prices.
The internet will hurt the record companies, that's why they hate it.
Pretty clever reasoning with that wineX-as-code-quality-checker argument. And it seems even today that games which "feel" better seem to be supported better under wineX. Another thing is that the makers of game _engines_ could have a big influence here, when they test their engine on wineX, it will do a lot for game compability, because all games based on that engine should be very easily portable. Though it's nicer when game engines are cross OS from the start.
But you are only theoretically right. _Real_ P2P doesn't meet the "classic" definition of broadcasting, but in effect, it is. This is, because it's a kind of "m:n casting". The - in retrospect - obvious, but also very ingenious idea of a "loosely codistributed" database (the clients peers) of filenames (or hashes) doesn't have the ubiquitousness of information that a chain of newsserver (databases) has, but it's is has a lot of other advantages, which outweigh that.
Well, maybe there is a need for that. Say, you have an application where you really don't want to, or easily can't (because of early design mistakes) adapt the application to big, effective cluster farms. Whatever, maybe a database, or a web server, which are quite fast, but also - because of that - quite quirky designed. Just examples...
This RDMA could help. Just push the problem one level deeper, to the OS and to hardware, at raw memory access. Let the OS try to figure out all the problems. Suddenly, you app _is_ (on paper) clusterable, with an (on paper) nearly linear scaling factor. Well it still sucks from the real performance POV, but you hope that your marketing department and carefully constructed example installations will save you, and in general technology progress (esp. with networks) will save your ass.
But we all now such a scheme wouldn't work, right?
No I understand they changed their mind, and I applaud them, but why not go the last step and use another, more obscure name if it's just a codename? Hell, look at the codenames Intel and MS choose.
Slashdot Mirror
Monty seemed to be saying that not having to consider crap like OS buffering would make implementing an ACID database a lot easier ...
But, they don't even know the OS they are programming to, so either you use O_SYNC when on posix, or you can't predict anything about VFS buffering. And even when using O_SYNC, you are in no way guaranted (when using pure IDE IIRC) in what order your commands will hit the disk, and when (on-disk cache).
So, I also have no idea w.t.f. the physical nature of the storage medium should have to do with ACID.
At least that way you don't wake up the next
...
morning not remembering anything and feeling
dirty...
The joys of literature
When you turn on this gadget, add your mailing list addresses to your white list. If you suddenly stop getting a list, go find out if they changed their sending address and add it to your white list.
Only to find out that the mailing list admin was fed up with umpteen "challenges" when he changed the sending adress and banned you for life.
Excellent idea, instant DOS attack:
The funny thing is, that a system like this might _drive_ spammers to use From: adresses which the deem more likely to be whitelisted (esp. since the possibility of whitelisting complete domains seems to be a nice feature at first).
Excellent idea, NOT!
Now one spam message creates a reply which has 100 fold the size of the average spam message, and, since the mail is forged anyway, goes nowhere.
Worse, if spammers forge valid adresses, one poor sob get's 5 Gigs of useless pictures of "validation emails" in his inbox instead of 1000 hatemails from lusers accusing him of spamming.
At least they should send pictures of naked supermodels with the confirmation secrets tatood on their butts.
And what happens if ReplyTo != From ?
Why would any terrorist worry about getting boxcutters past airport security now when they could dump an assload of ricin into a big city's reservoir and watch hundreds of thousands of people croak?
Erm, obviously they decided to do the plane thing, although there were _no_ homeland security stuff and _no_ cameras installed at that time.
People, face it, there are a million things dedicated evil people can do and _nothing_ can really prevent it. America is _not_ at war (at least there's no war at its home turf - ask any nation who really was hit by war), and all this stuff does is to cause mass paranoia without gaining anything.
Oh wait, there are nice companies gaining from paranoia, but that's it.
OTOH, there's an interesting short story in one of Hofstaedter's (spelling right? to lazy to look up) books where one guy build mechanical "bugs" which could express very simple "emotions" - i.e. making some pet like noises, making noises of fear, crawinling around etc. - and he asks a visitor to smash one bug with a hammer. The story describes how this man wasn't emotionally able to do that, because he developed feelings for this bug.
(IIRC the real story is somewhat more involved, but you get the idea).
I bet if you decorated an "intelligent" AI with some emotional dressing, you could significantly lower the barrier to accept it as "intelligent".
Shows how deeply involved the human perception of not only intelligence, but life in general is.
But doesn't that example illustrate rather a human "deficiency" than one from a machine?
I mean the deficiency that there is far more "computation" going on in our brain than we consciously notice. I don't think that this "right feeling" was the result of something supernatural going on.
To add another example. Richard Feynman writes in one of his books about how he listened to a conversation between to students of mathematics(at MIT or so) where one tried to explain some mathematical concept to the other.
Feynman described that after a long while and much intense explaining, finally the other student "got it", and said something along the lines of "Oh! YES, THAT'S TRIVIAL!"
Feynman goes on to make fun of mathematicians by proposing that mathematicians only understand trivial problems, because anything they have already understood is declared trivial by them.
This is a bit extreme, but it decribes exactly the notion some AI critics seem to have when judging AI advances.
1) Make a computer with true free will. Let's see AI do something it wasn't originally designed to do because it wants to.
The problem is that philosphers/psycologists/neurologists/genome scientists are debating since centuries if and in what magnitude this sentence is true when you replace computer and AI with man.
Exactly.
And since philophers and other scientists are not dumb, they even go to the root of the problem in advance.
Searle's chinese room theory is a prime example for this IMO, you could use this reasoning to rip apart _anything_ which will come out of AI science. Note that I don't think that AI today is able to resemble anything which will really master a turing test, but even if it did, the chinese room argument will offer a way to argue that the computer "doesn't really understand what he is talking."
I for one think that this argument is silly, because it relies on our inability (today) to deconstruct the brain in the same way we could with a software.
If we could, we'd find out that the different "parts" of our brain don't really "understand" what they are doing - in the same way.
Oh, and cars aren't really able to go forward, because an engine is not able to move on itself.
Take a step back and look closely at the submitters name and the reference website's name.
I doubt he would misread his own article.
The computer graphics world lost a brilliant man when Kai Krause retired.
He didn't really retire. In case you're interested, you can find the website of his project at http://www.byteburg.de/
Go to the german version, there is also english content.
Hmm, looks like the game publishers should just start to put their demos on the big p2p networks.
That could be the death of fileplanet et al.
Found the server, read about it here
http://www.chromium.com/x15tech.html
There is at least on userspace daemon which reach the same performance/scalability on linux (commercial, I don't remember its name, the creator posted on linux-kernel about it). The point is that when implementing tux, they made a lot of general improvements in the kernel. These are the reason for tux' performance, not the fact that it is in kernel space. Therefore the spec numbers of tux give far more substance to the scalability debate than 5 years old hearsay.
FreeBSD 5.0 seems superior to Linux in the role of webserver when it comes to scaling, ...
This myth has to die or someone has to prove it, with recent versions of linux and bsd. There have been so many advances with linux (for instance after the mindcraft incident etc.) that I doubt it's true.
The current advances in linux can clearly be seen over at spec.org, linux with the tux webserver scales nearly linearly to 8 procs.
I didn't want to write too much.
...
First, a player like Microsoft certainly can buy enough marketing power, and second, if the internet gets a more important distribution channel for a good, internet specific marketing will get more important for that good.
A lot of music marketing consists of paying shops for putting albums in prominent position, shipping cds in as much shops you can, paying shops for displaying your cd in their advertisements etc, which is all related to the classical distribution methods.
Oh, and that thing about taking risks
It's not that artists don't want to take the risk, they just can't afford it. If they had the money (or could get a credit), I'm sure most wanted to take the risk under the same circumstances (e.g. marketing power), because for all I read, making a contract with label is mainly an very expensive loan.
Wrong. The hardware and support costs can be amortized on a huge scale. How many hosting centers is this? How many distribution centers are CDs shipped from? Do the math, its not even close. Record companies are going to make a killing on cost davings alone if this pans out.
I think and you are wrong, and that you are wrong for the same reason record companies are trying to get draconian with copy protection.
If this takes off and record companies enter the game in a big way, it will take off big - very big. So big that it has the potential to badly damage the classic distribution chains. Music is the optimal good for distribution over the net in the state it has today (average bandwidth for the end user).
Fast dsl/cable connections make the act of purchasing and downloading music in a compressed format unpremeditated buying.
After online sales getting a significant share of the total revenue, there's suddenly a very low barrier of entry for anyone for this business.
Why do musicians sign their soul to big music companies?
Because they are the only one offering the the things they need (or believe they need in case of the first):
- marketing power
- logistics (they can make an album appear in every shop on the planet)
It's clear the internet solves the logistics, and this is IMO the biggest hinderance for newcomers. It also could raise the absolute number of sales (unpremeditated buying etc.).
But also completely new competitors could emerge, or artists might consider handling their own sales, which all will eventually drive down prices.
The internet will hurt the record companies, that's why they hate it.
Pretty clever reasoning with that wineX-as-code-quality-checker argument. And it seems even today that games which "feel" better seem to be supported better under wineX.
Another thing is that the makers of game _engines_ could have a big influence here, when they test their engine on wineX, it will do a lot for game compability, because all games based on that engine should be very easily portable. Though it's nicer when game engines are cross OS from the start.
But you are only theoretically right. _Real_ P2P doesn't meet the "classic" definition of broadcasting, but in effect, it is. This is, because it's a kind of "m:n casting". The - in retrospect - obvious, but also very ingenious idea of a "loosely codistributed" database (the clients peers) of filenames (or hashes) doesn't have the ubiquitousness of information that a chain of newsserver (databases) has, but it's is has a lot of other advantages, which outweigh that.
Well, maybe there is a need for that. Say, you have an application where you really don't want to, or easily can't (because of early design mistakes) adapt the application to big, effective cluster farms. Whatever, maybe a database, or a web server, which are quite fast, but also - because of that - quite quirky designed. Just examples ...
This RDMA could help. Just push the problem one level deeper, to the OS and to hardware, at raw memory access. Let the OS try to figure out all the problems. Suddenly, you app _is_ (on paper) clusterable, with an (on paper) nearly linear scaling factor. Well it still sucks from the real performance POV, but you hope that your marketing department and carefully constructed example installations will save you, and in general technology progress (esp. with networks) will save your ass.
But we all now such a scheme wouldn't work, right?
Well, yeah, that's what I was onto, just didn't want to say it as directly as you did ;).
You are refering to my ps right?
No I understand they changed their mind, and I applaud them, but why not go the last step and use another, more obscure name if it's just a codename? Hell, look at the codenames Intel and MS choose.