Freenet does not push data around. Data moves around the network in response to requests, ie it is pulled, and in a much more efficient manner than Gnutella (no broadcasting).
You will not find dozens of identical versions of a file on Freenet, because when a document is inserted into Freenet its key is generated by hashing the data. That means that two identical copies of a document will have identical keys, and thus they will never be stored redundantly on the same node.
The amount of data stored is, as far as I can see, irrelevant. I don't understand where your hundreds of gigabytes/terabytes argument is coming from at all. Unlike Gnutella, Freenet should be able to scale to thousands or millions of nodes without breaking up into isolated islands. Maybe your concern about the size of "all the popular music" is based on the idea that Freenet is meant to be a Napster replacement? It's not. But if people want to use Freenet to trade MP3s, and they're willing to provide the disk space, it's certainly up to the task.
Also, it seems to me that any network in which a specific document can eventually be tracked to a single IP address is insecure. While it can never be shut down, per se, anyone who is doing anything that make *make* someone want to shut it down can still be found (at least until the mibs knock at their door).
Freenet takes this vulnerability into account in two ways:
First, you don't send out a query to a central server and get back an IP address, like you do with DNS or Napster. You send out a query to a neighbouring Freenet node and get back a document (or a failure notice). You can't tell whether the document came from the node to which you sent the request, or from one of your other neighbours, or from another node whose IP address you've never heard of. All you know is that it passed through the node to which you sent the request. But since it may be stored somewhere else, that knowledge doesn't help you to censor the document.
However, if every request was routed through dozens of nodes before the requested document was found, and if the document had to be passed back through as many nodes, Freenet would be extremely slow. Freenet has two clever properties to avoid this: "route compression" and "document clustering".
Route compression
When a node responds to a request by returning a document (call it document X), it can set the DataSource field of X's metadata to indicate the node from which X was obtained. (This may not be the actual source, but it cuts out some of the middlemen.) The requester uses this address when looking for documents with keys lexographically close to the key of X, so the intermediate nodes are skipped in any future requests. Effectively, a new edge in the Freenet graph is created, linking the requester to the node listed in the DataSource field.
Document clustering
This works almost like a neural network "learning" the fastest route to a document. If a node successfully obtains document X from node Y, it will forward subsequent requests with keys similar to the key of X to node Y. Over time, Y's cache will fill up with documents having similar keys, clustered around the key of X. This will increase the likelihood of success of requests for keys inside the cluster, and increase the likelihood of failure of requests for keys outside the cluster (as unrequested documents are squeezed out of the cache). The sets of documents cached by Y while serving requests for each of its neigbours will start to overlap, and the sets will eventually become identical - Y will become an "authority" on a particular part of the keyspace. Over the network as a whole, documents with lexicographically close keys (note that this has nothing to do with the contents of the documents) will tend to cluster together on the same node. Thus over time, requests will be routed more and more efficiently as the "guesses" made by nodes about which neighbour to forward a request to become more and more accurate.
The second way that Freenet avoids censorship is through replication. Every node involved in tunnelling a request caches the requested document as it passes through. This not only allows documents to cluster, it also ensures that popular documents are replicated across the network while unpopular documents eventually disappear. Any attempt to censor a document by taking advantage of route compression to discover the IP address of the node storing the document is bound to fail, because you have to request the document dozens of times to be sure you have found the true source of the document, and by that time you have succeeded in spreading the document all over Freenet. It's a very elegant self-balancing mechanism.
One man's moral stand is another man's censorship
on
Freenet 0.3 Released
·
· Score: 2
How can I ensure that my machine is not involved in the trafficking of content that I don't support?
You can't, whether you run a Freenet node or not.
If your machine is connected to the internet, it's probably forwarding packets full of kiddie porn, white supremacist bullshit and a thousand other flavours of poison all the time. If your machine uses a dialup connection, you're just paying someone else to forward those packets. The only change when you start running a Freenet node is how long you sit on the data.
It's important that Freenet node operators can't (easily) filter the contents of the datastore, because if they could, a node operator could be held responsible for the contents of the datastore if he didn't filter it. Would you want to be held accountable for the contents of every packet passing through your machine (or your ISP's machine)?
How will they include an inaudible watermark, yet ensure that it is reproduced by all playback equipment? There are three possible ways to watermark an audio signal:
The watermark frequencies could be outside the audible range (20 Hz - 20 kHz). But most audio equipment filters out, or fails to reproduce, inaudible frequencies.
The watermark frequencies could be inside the audible range, disguised or masked by the music. But MP3 compression works by removing inaudible, masked sounds.
The watermark could be encoded steganographically. But resampling will alter the least significant bits of the recording and destroy steganographic information.
What's to stop me connecting the analogue outputs of my soundcard to the analogue inputs of another soundcard (to remove out-of-band frequencies and steganographic data), then making an MP3 of the result (to remove masked tones)? Only the fact that the hardware and software that would allow me to do so (a Linux PC with two Soundblaster 16s and a copy of BladeEnc) will be illegal.
SDMI cannot succeed technologically, and the record companies know this. The technology only exists to provide an excuse for legal restrictions on hardware and software. Any hardware or software that is not "SDMI-compliant" will be branded a "circumvention device" and banned in the USA. Any countries that fail to follow the US's lead will be branded "havens for piracy".
SDMI aims to stop bootlegging by placing a unique, inaudible watermark into every copy of a song. That way, when the RIAA finds an MP3 of the song on Freenet, they can identify the original copy from which the bootleg copy was made. But how does this bring them any closer to prosecuting the bootlegger? Before they can do that, they will need proof that the individual in question downloaded a particular watermarked copy.
How are they going to stop me from buying songs as Chuck U. Farley, then bootlegging them to my heart's content? They will require me to pay by credit card. My credit card will become my proof of identity - the proof that I exist in the real world, at a known address, with a real door that can be kicked down. And if I lose my credit card, and my neighbour uses it to buy songs online, songs which he subsequently puts on Freenet? Oops, I'm liable. The credit card company might pay your bill when your card is stolen, but they won't go to jail for you.
We need an anonymous micro-payment system right now.
So they'll release (and hype) the technology under the impression that it's secure, and three months later, when the record companies have poured billions into the technology, somebody will discover a flaw. What a shame.
By the time you can afford a high-definition TV and a high-fidelity sound system, your sight and hearing will have degraded to the point where everything sounds like an MP3 and looks like an MP4.
Akamai is awesome, but it would be nice to have an open-source, open-network way to implement this at the ISP level. Does anyone know of any such effort underway already?
It seems that the KDE developers, no matter what they do, no matter what good intentions they hold, always gets bashed by the GNU/GNOME/RMS camp.
GNU hackers fear the unfamiliar - years of subtle incompatibilities between dozens of versions of Unix (and now dozens of distros of Linux) have made them twitchy and distrustful of anything except strict GNU orthodoxy. An acronym that doesn't start with G immediately sets mental alarm bells ringing:
"This isn't one of our projects... it must come from outside. It could be contaminated! Un-free heretic! Burn the witch! This is a local shop for local people, etc"
KDE could stop the flamewars tomorrow if they renamed their project GDE.
I think their point is that if they can be sued for making it possible to download MP3s, then they can sue Gnutella for making it possible to download MP3s. They are deliberately acting in an absurd way to highlight the absurdity of their own case.
While burning the flag may be speech, setting light to someone's house with a burning flag is not. You have to draw a line between actions which are purely symbolic, and actions which have real physical consequences. In US law I believe this line is drawn between description of an action ("Gates must die") and incitement to that action ("go and murder Gates"). Source code uncomfortably straddles this line, since it is both expressive and imperative; it both describes an action (to the hacker) and commands it (to the computer).
I would assume that a natural-language description of source code, which cannot be trivially interpreted as a series of commands but which still expresses the commands which could be executed, would be covered by the First Amendment.
The DeCSS source code is available in steganographic form here and here, with instructions here. Or you can grab the tarball here. Slashdot is now guilty of linking to the source code and can be shut down. Well actually, it's not quite that simple. The ruling contains this protection for webmasters:
Accordingly, there may be no injunction against, nor liability for, linking to a site containing circumvention technology, the offering of which is unlawful under the DMCA, absent clear and convincing evidence that those responsible for the link (a) know at the relevant time that the offending material is on the linked-to site, (b) know that it is circumvention technology that may not lawfully be offered, and (c) create or maintain the link for the purpose of disseminating that technology.
The problem for Slashdot is, who are "those responsible for the link"? Does it mean me for submitting the link? Or CmdrTaco for allowing it to be submitted?
Does Slashdot have a responsibility to prevent me from submitting links to the source code? To enforce that interpretation you'd have to shut down every news site and message board in the US.
Well maybe I'm responsible for the link. But I'm in the UK, outside the jurisdiction of Judge Kaplan, the Supreme Court and the DMCA. They can't stop me submitting comments to American sites.
The DeCSS source code is available in steganographic form here and here, with instructions here. Or you can grab the tarball here. Slashdot is now guilty of linking to the source code and can be shut down. Well actually, it's not quite that simple. The ruling contains this protection for webmasters:
Accordingly, there may be no injunction against, nor liability for, linking to a site containing circumvention technology, the offering of which is unlawful under the DMCA, absent clear and convincing evidence that those responsible for the link (a) know at the relevant time that the offending material is on the linked-to site, (b) know that it is circumvention technology that may not lawfully be offered, and (c) create or maintain the link for the purpose of disseminating that technology.
The problem for Slashdot is, who are "those responsible for the link"? Does it mean me for submitting the link? Or CmdrTaco for allowing it to be submitted?
Does Slashdot have a responsibility to prevent me from submitting links to the source code? To enforce that interpretation you'd have to shut down every news site and message board in the US.
Well maybe I'm responsible for the link. But I'm in the UK, outside the jurisdiction of Judge Kaplan, the Supreme Court and the DMCA. They can't stop me submitting comments to American sites.
1. Is it illegal to run a website which links to the DeCSS source code if visitors must click through a link on the front page labelled "The following page may contain links to the DeCSS source code. Only click here if you wish to be exposed to these links"?
2. Is it illegal to run a website which links to the DeCSS source code if visitors must type "DeCSS source" into a text box on the front page to view the links?
3. Is it illegal to run a search engine which links to the DeCSS source code if visitors must type "DeCSS source" into a text box on the front page to view the links?
...to think that Intel would pose a threat to their high end server business just because of OS is ludicrous.
Quite. That's why I said "Sun can charge a premium for Solaris workstations because they're interoperable with its big servers". I did not say people would be replacing their Sun servers with PCs - that would be insane. But people who buy "Sun everything" because they want an interoperable setup may now be tempted to mix Sun servers with Intel desktop machines. As my university's CS department is already doing, in fact.
Who's going to run Linux on an expensive workstation when they can run exactly the same software on a cheap PC? At the moment Sun can charge a premium for Solaris workstations because they're interoperable with its big servers. If it releases a version of Linux which is equally interoperable, who's going to run it on SPARC hardware? Of course Sun can still lock people in by refusing to support Linux on any platform except SPARC, but will customers stand for that?
You can split a large file into pieces and insert them separately. Then insert a list of the pieces as another file. Voila - large file support. The 100k limit is almost certainly there to avoid attacks which flood the system using bottomless data streams, or use large files to diplace a disproportionate number of smaller files. A similar limit will be used on Freenet, for similar reasons. It doesn't drastically hurt the system's ability to support freedom of speech; it protects it.
The 100k restriction is almost certainly there for technical reasons, not to keep out MP3s. After all, you can just split your MP3 into multiple parts and insert them separately. Freenet will probably have limits on file size, and they are *not* interested in inhibiting your free speech. Limiting the size of files makes things like caching and stream handling easier, because you don't have to chug down endless data streams sent by malicious nodes - you just cut them off after the first 100k.
Slashdot Mirror
Freenet does not push data around. Data moves around the network in response to requests, ie it is pulled, and in a much more efficient manner than Gnutella (no broadcasting).
You will not find dozens of identical versions of a file on Freenet, because when a document is inserted into Freenet its key is generated by hashing the data. That means that two identical copies of a document will have identical keys, and thus they will never be stored redundantly on the same node.
The amount of data stored is, as far as I can see, irrelevant. I don't understand where your hundreds of gigabytes/terabytes argument is coming from at all. Unlike Gnutella, Freenet should be able to scale to thousands or millions of nodes without breaking up into isolated islands. Maybe your concern about the size of "all the popular music" is based on the idea that Freenet is meant to be a Napster replacement? It's not. But if people want to use Freenet to trade MP3s, and they're willing to provide the disk space, it's certainly up to the task.
Freenet takes this vulnerability into account in two ways:
First, you don't send out a query to a central server and get back an IP address, like you do with DNS or Napster. You send out a query to a neighbouring Freenet node and get back a document (or a failure notice). You can't tell whether the document came from the node to which you sent the request, or from one of your other neighbours, or from another node whose IP address you've never heard of. All you know is that it passed through the node to which you sent the request. But since it may be stored somewhere else, that knowledge doesn't help you to censor the document.
However, if every request was routed through dozens of nodes before the requested document was found, and if the document had to be passed back through as many nodes, Freenet would be extremely slow. Freenet has two clever properties to avoid this: "route compression" and "document clustering".
Route compression
When a node responds to a request by returning a document (call it document X), it can set the DataSource field of X's metadata to indicate the node from which X was obtained. (This may not be the actual source, but it cuts out some of the middlemen.) The requester uses this address when looking for documents with keys lexographically close to the key of X, so the intermediate nodes are skipped in any future requests. Effectively, a new edge in the Freenet graph is created, linking the requester to the node listed in the DataSource field.
Document clustering
This works almost like a neural network "learning" the fastest route to a document. If a node successfully obtains document X from node Y, it will forward subsequent requests with keys similar to the key of X to node Y. Over time, Y's cache will fill up with documents having similar keys, clustered around the key of X. This will increase the likelihood of success of requests for keys inside the cluster, and increase the likelihood of failure of requests for keys outside the cluster (as unrequested documents are squeezed out of the cache). The sets of documents cached by Y while serving requests for each of its neigbours will start to overlap, and the sets will eventually become identical - Y will become an "authority" on a particular part of the keyspace. Over the network as a whole, documents with lexicographically close keys (note that this has nothing to do with the contents of the documents) will tend to cluster together on the same node. Thus over time, requests will be routed more and more efficiently as the "guesses" made by nodes about which neighbour to forward a request to become more and more accurate.
The second way that Freenet avoids censorship is through replication. Every node involved in tunnelling a request caches the requested document as it passes through. This not only allows documents to cluster, it also ensures that popular documents are replicated across the network while unpopular documents eventually disappear. Any attempt to censor a document by taking advantage of route compression to discover the IP address of the node storing the document is bound to fail, because you have to request the document dozens of times to be sure you have found the true source of the document, and by that time you have succeeded in spreading the document all over Freenet. It's a very elegant self-balancing mechanism.
You can't, whether you run a Freenet node or not.
If your machine is connected to the internet, it's probably forwarding packets full of kiddie porn, white supremacist bullshit and a thousand other flavours of poison all the time. If your machine uses a dialup connection, you're just paying someone else to forward those packets. The only change when you start running a Freenet node is how long you sit on the data.
It's important that Freenet node operators can't (easily) filter the contents of the datastore, because if they could, a node operator could be held responsible for the contents of the datastore if he didn't filter it. Would you want to be held accountable for the contents of every packet passing through your machine (or your ISP's machine)?
The watermark frequencies could be outside the audible range (20 Hz - 20 kHz). But most audio equipment filters out, or fails to reproduce, inaudible frequencies.
The watermark frequencies could be inside the audible range, disguised or masked by the music. But MP3 compression works by removing inaudible, masked sounds.
The watermark could be encoded steganographically. But resampling will alter the least significant bits of the recording and destroy steganographic information.
What's to stop me connecting the analogue outputs of my soundcard to the analogue inputs of another soundcard (to remove out-of-band frequencies and steganographic data), then making an MP3 of the result (to remove masked tones)? Only the fact that the hardware and software that would allow me to do so (a Linux PC with two Soundblaster 16s and a copy of BladeEnc) will be illegal.
SDMI cannot succeed technologically, and the record companies know this. The technology only exists to provide an excuse for legal restrictions on hardware and software. Any hardware or software that is not "SDMI-compliant" will be branded a "circumvention device" and banned in the USA. Any countries that fail to follow the US's lead will be branded "havens for piracy".
How are they going to stop me from buying songs as Chuck U. Farley, then bootlegging them to my heart's content? They will require me to pay by credit card. My credit card will become my proof of identity - the proof that I exist in the real world, at a known address, with a real door that can be kicked down. And if I lose my credit card, and my neighbour uses it to buy songs online, songs which he subsequently puts on Freenet? Oops, I'm liable. The credit card company might pay your bill when your card is stolen, but they won't go to jail for you.
We need an anonymous micro-payment system right now.
Um, I think that was Alasteir Crowley. :)
So they'll release (and hype) the technology under the impression that it's secure, and three months later, when the record companies have poured billions into the technology, somebody will discover a flaw. What a shame.
By the time you can afford a high-definition TV and a high-fidelity sound system, your sight and hearing will have degraded to the point where everything sounds like an MP3 and looks like an MP4.
Yes.
You can use up the space even quicker (and thus cost the RIAA members even more money) if you don't encode them.
GNU hackers fear the unfamiliar - years of subtle incompatibilities between dozens of versions of Unix (and now dozens of distros of Linux) have made them twitchy and distrustful of anything except strict GNU orthodoxy. An acronym that doesn't start with G immediately sets mental alarm bells ringing:
"This isn't one of our projects... it must come from outside. It could be contaminated! Un-free heretic! Burn the witch! This is a local shop for local people, etc"
KDE could stop the flamewars tomorrow if they renamed their project GDE.
I think their point is that if they can be sued for making it possible to download MP3s, then they can sue Gnutella for making it possible to download MP3s. They are deliberately acting in an absurd way to highlight the absurdity of their own case.
I would assume that a natural-language description of source code, which cannot be trivially interpreted as a series of commands but which still expresses the commands which could be executed, would be covered by the First Amendment.
Accordingly, there may be no injunction against, nor liability for, linking to a site containing circumvention technology, the offering of which is unlawful under the DMCA, absent clear and convincing evidence that those responsible for the link (a) know at the relevant time that the offending material is on the linked-to site, (b) know that it is circumvention technology that may not lawfully be offered, and (c) create or maintain the link for the purpose of disseminating that technology.
The problem for Slashdot is, who are "those responsible for the link"? Does it mean me for submitting the link? Or CmdrTaco for allowing it to be submitted?
Does Slashdot have a responsibility to prevent me from submitting links to the source code? To enforce that interpretation you'd have to shut down every news site and message board in the US.
Well maybe I'm responsible for the link. But I'm in the UK, outside the jurisdiction of Judge Kaplan, the Supreme Court and the DMCA. They can't stop me submitting comments to American sites.
How's this ban on linking supposed to work again?
Accordingly, there may be no injunction against, nor liability for, linking to a site containing circumvention technology, the offering of which is unlawful under the DMCA, absent clear and convincing evidence that those responsible for the link (a) know at the relevant time that the offending material is on the linked-to site, (b) know that it is circumvention technology that may not lawfully be offered, and (c) create or maintain the link for the purpose of disseminating that technology.
The problem for Slashdot is, who are "those responsible for the link"? Does it mean me for submitting the link? Or CmdrTaco for allowing it to be submitted?
Does Slashdot have a responsibility to prevent me from submitting links to the source code? To enforce that interpretation you'd have to shut down every news site and message board in the US.
Well maybe I'm responsible for the link. But I'm in the UK, outside the jurisdiction of Judge Kaplan, the Supreme Court and the DMCA. They can't stop me submitting comments to American sites.
How's this ban on linking supposed to work again?
1. Is it illegal to run a website which links to the DeCSS source code if visitors must click through a link on the front page labelled "The following page may contain links to the DeCSS source code. Only click here if you wish to be exposed to these links"?
2. Is it illegal to run a website which links to the DeCSS source code if visitors must type "DeCSS source" into a text box on the front page to view the links?
3. Is it illegal to run a search engine which links to the DeCSS source code if visitors must type "DeCSS source" into a text box on the front page to view the links?
Guns don't kill people. ED-209 kills people.
Just goes to show how much things have changed...
Quite. That's why I said "Sun can charge a premium for Solaris workstations because they're interoperable with its big servers". I did not say people would be replacing their Sun servers with PCs - that would be insane. But people who buy "Sun everything" because they want an interoperable setup may now be tempted to mix Sun servers with Intel desktop machines. As my university's CS department is already doing, in fact.
Who's going to run Linux on an expensive workstation when they can run exactly the same software on a cheap PC? At the moment Sun can charge a premium for Solaris workstations because they're interoperable with its big servers. If it releases a version of Linux which is equally interoperable, who's going to run it on SPARC hardware? Of course Sun can still lock people in by refusing to support Linux on any platform except SPARC, but will customers stand for that?
Now we can make some money from GPL'd software: give the source code away for free, and charge for the comments.
You can split a large file into pieces and insert them separately. Then insert a list of the pieces as another file. Voila - large file support. The 100k limit is almost certainly there to avoid attacks which flood the system using bottomless data streams, or use large files to diplace a disproportionate number of smaller files. A similar limit will be used on Freenet, for similar reasons. It doesn't drastically hurt the system's ability to support freedom of speech; it protects it.
The 100k restriction is almost certainly there for technical reasons, not to keep out MP3s. After all, you can just split your MP3 into multiple parts and insert them separately. Freenet will probably have limits on file size, and they are *not* interested in inhibiting your free speech. Limiting the size of files makes things like caching and stream handling easier, because you don't have to chug down endless data streams sent by malicious nodes - you just cut them off after the first 100k.
Thanks for the link! I have a copy of Oersted, but that's the only recording of his I could get hold of.
Nic Endo