Re:cry me a river you CRIMINAL
on
DMCA 2, Freedom 0
·
· Score: 2, Interesting
The DMCA only covers copyrighted material, not material which is in the public domain.
This is not true. The anticircumvention portions talk about devices and technologies which are used to protect access to copyrighted works--- it is the technology itself which is illegal. It doesn't matter whether DeCSS is used to access copyrighted or non-copyrighted works, the "device" itself is still illegal. (The judge in the 2600 case said that this issue wasn't yet "ripe" because nobody's tried to prevent access to non-copyrighted works yet.)
Interestingly, a tool which is legal today (say, a DES cracker) might become illegal tomorrow if somebody starts using a content protection scheme which it defeats.
The author of the article isn't particularly careful about that distinction...
It's not clear that this _is_ theft of service, anyway (although many cable modem agreements state that you can't share the service with others.) Somebody is paying the bill for the pipe--- I don't believe that the ISP is owed money per device rather than for bandwdith. Would it still be "stealing" if a web proxy was used rather than NAT?
This only helps if the document is only in Google's cache. If it isn't, the attacker can just follow the link and read the original document.
Excluding entire documents that happen to match the checksum from the search results would work, but it'd be interesting to see how many false positives would result...
Not just the huddled masses. How many IP addresses do you have memorized? Even if you can still read slashdot, how much work will even the technically elite get done without email, which is very DNS-dependent? IP routing may go on, but my view is that DNS is a pretty central part of how the Internet works... Attacks on DNS are very scary because they interfere with the normal channels for responding to those attacks.
I'm working on a paper on this very topic (attacks against DNS infrastructure and how to defend against it). I don't think existing work on DDoS attacks is sufficient for the scale of assaults we may see in the future...
By default, tinydns does not hand out referrals to questions it is asked about zones it does not control. I believe that this violates the spirt of the RFCs, if not the letter.
Please indicate where do you think that this breaks the RFCs.
RFC 1034, section 4.3.1 says that the legal responses (for a non-recursive query) are:
An authoritative name error indicating that the name does not exist.
A temporary error indication.
Soome combination of: RRs that answer the question, together with an indication
whether the data comes from a zone or is cached. A referral to name servers which have zones which are closer ancestors to the name than the server sending the reply.
RRs that the name server thinks will prove useful to the requester.
Even though giving back a "temporary error" is technically legal, it is not within the "spirit" of the specification to choose to do so always. Further, the algorithm in section 4.3.2 states:
... b If a match would take us out of the authoritative data,
we have a referral. This happens when we encounter a node with NS RRs marking cuts along the bottom of a zone.
Copy the NS RRs for the subzone into the authority section of the reply. Put whatever addresses are available into the additional section, using glue RRs if the addresses are not available from authoritative data or the cache. Go to step 4.
That is the idea. However, the RFC (2136) only requires that the changes be committed to stable storage--- and explicitly state that this may involve writing _just_ the changes, since writing out a huge text file could be expensive.
Horowitz has been writing columns for Salon since February 10, 1997--- hardly "recently". On the other hand, he is a loon who probably does little to change anybody's mind.
Heck, compared to the local San Francisco weeklies, Salon is the essence of evenhandedness. Do you have any online alternatives to suggest? I can't stand Slate (nor do I have any love for the company which owns it...)
Thanks for your coherent and insightful answer. I hadn't thought about Forth in quite that way, although it's been a long time since I played around with it. One could view unix piping as a concatenative programming language by defining "tokens" correctly...
Naturally, this behavior doesn't come without cost. With Forth (and Joy) it seems to me the cost of not having any arguments is that you can't do any type checking. Indeed, wouldn't any large Forth project have to document what "arguments" a word expects? I'm not convinced us poor humans can deal with all the power at our fingertips.:)
I think that one _could_ probably build a type-inference system into Joy. Basically all types would be parameterized--- rather than "int" it would be "stack of type X with int on top".
I like it, too. But, I think that the biggest problem is how close the Klingon homeworld is. If it's 4 days (!) away at warp 4, then it's just hours away with "modern" (TNG) warp technology.
This makes much of the future inconsistent. For example, why do no Klingon ships show up to battle the Borg? How come Earth never has anybody but the Enterprise to defend it from threat X--- isn't there any defensive fleet? (more of a problem with TOS.)
I'm also kinda annoyed that the doctor has to be a brand-new species--- could have used one of the throwaways from earlier shows. (Or did they?)
I think the benefits of Forth in terms of "thinking different" could just as easily be applied to other languages such as Perl, Scheme or ML. Each language has its own idioms and ways of approaching problems. Most of the points I've heard in favor of Forth boil down to "it's small", "it uses postfix", and "it forces you to write small functions." None of these seem particularly convincing to me.
I really can't understand what's so appealing about postfix notation as opposed to prefix or infix. (Despite what Chuck Moore says, most modern languages are actually a combination of the two, which does give Forth some edge in purity, especially for people who don't like parentheses.)
How would he then tell people how to look up the information on Freenet? If he already has an anonymous communication channel with which to distribute a Freenet key, then he could use that directly to distribute the information. If no such channel is available, then posting the key in a traceable manner will still expose him.
Someone could just try guessing random related names, of course, but I (or Intel) could just as easily guess what people would try and preemptively put information under those names as well...
I think there are enough users (in terms of Linux desktop installations) but not enough people who use exclusively non-Windows OSes. Anybody who dual-boots is going to be hard to sell to.
I gave up using Windows entirely (at home) about the time of my last hard drive crash, so I was naturally excited about Loki, and own two of their games. (And will probably buy Kohan if it's still available...) But I'm not a typical game-player--- and I can understand how someone who spends more time would have a Windows partition available and make use of it. Heck, I'm still tempted to buy Black and White and try to play it on my work laptop...
*sigh* I really hope I don't break my promise to give no more of my money to Bill G...
I've had @Home in the bay area for about 6 months, since moving to a new apartment meant I couldn't keep DSL. It's been an incredibly frustrating experience, with entire weeks in which we saw downtime for large periods of time every day. Our connectivity still has a tendency to migrate into a 70%-packet loss, 1000ms+ latency state and stay there for a while.
Not only that, but everyone in the apartment is frustrated by ATT @Home's tendency to treat us like idiots during every stage of our relationship with them, from the initial installation to complaints to customer service.
I'd switch back to DSL in a heartbeat (now that it's available in my area), except I really can't afford the extra cost to get PacBell's "Enhanced" service, nor do I want to fool around with the evil dynamic-IP PPP-over-Ethernet garbage they give with the Basic service these days.
Given that @Home needs more funding in order to stay afloat past the end of the year, according to an article yesterday in the San Jose Mercury Times, I wouldn't be surprised at all if their assets were acquired or if ATT switched partners.
The U.S. government is making itself look really bad with this incident, particularly at a time when it's trying to get U.S. scholars released from Chinese prisons. I'm surprised Russia and China haven't raised a stink about this, given that jurisdiction is a little shaky in this case--- has anybody seen articles about any reaction from the Russion government?
I'm very tempted to write my congresscritter and point out that arresting foreign citizens for "crimes" committed in their native land (where they are perfectly legal) is exactly the human rights violation we're protesting when China does it to U.S. citizens. Of course, the U.S. government has no problem being hypocritical anyway, but it still must look bad to the rest of the world...
However, these aren't the only parties--- I could easily see a judge dismissing the RIAA and SDMI from the case, but the complaint lists the U.S. Government (well, at least the AG) as targets as well. As others have pointed out, there are criminal penalties to be careful of, too. Given the recent arrest, I'd say there's a pretty "affirmative fear" there.
Does anyone know what the procedure would be if some of the "defendants" were dropped? Would the EFF have to refile or just amend their complaint?
I agree. I had both an Erector set and Legos as a kid. Lego won hands down; I wasn't interested in building bridges and cranes and other mechanical engineering stuff. I was interested in building spaceships, robots, and supercomputer installations. And although I kind of liked the chapters on organic chemistry in high school, no molecule modelling set held my attention for long.
(I don't like the newer Lego stuff with all the funny shaped pieces. I haven't decided whether it's because you really can't build anything other than the model they suggest with them, or whether I'm too old and dumb to come up with creative uses.)
I think you are definining Computer Science too narrowly. CS includes not only the theory/science of algorithms but also the more engineering disciplines of computer architecture and software design. A good CS grad should be able to design a system, write code, and prove theorems.
What one should get from a CS degree is understanding, which necessarily involves almost every level of a system. Any physicist can learn to program, but most don't understand why, say, the order of nested loops matters when jumping through an array. Some theoreticians come up with elegant algorithms that are beaten by simpler ones which use a smaller working set. Machine language programmers sometimes lose sight of the fact that a better data structure would do more for performance than any pipeline optimization they can come up with. And C programmers can be baffled by the fact that their distributed system breaks despite the lack of obvious bugs on any single machine.
So, I agree that there are no "universal truths" to be discovered by assembly language programming. But you would go too far to say that there isn't any insight to be gained by being familiar enough with that level to do so--- after all, even in his new revision of the Art of Computer Programming, Knuth decided it was still worth expressing things in an assembly-like language to truly understand what the real performance costs were.
Now, you may have been trying to make the point that all these are really "software engineering" concerns, but I disagree that computer science is only the mathematical parts--- without the practical angle, CS is only so much hot air and might as well stay part of the math department.
Re:But did Kubrick write the meta-science?
on
Review: A.I.
·
· Score: 1
A bigger problem is--- why does spinach break David, but being under water does not? Could David have been drinking all the water he wanted to?
I think this is a very clever design, and nicely illustrates how to use Freenet's signed-subspace keys to avoid interference from a third party. I do believe, however, that it suffers from a couple problems inherent in the design of Freenet--- although only the first looks like a showstopper to me.
Massive contention on game starts. If you restrict yourself to using just Freenet, the only way to find a game is to start searching "today's" namespace. This is a very expensive operation on Freenet; in addition, the amount of time needed to start a game will grow linearly with the number of people interested in playing it.
Perhaps a better algorithm for searching could help (for example, finding the "next" number on game creation by doing a search for 1, 2, 4, 8, etc., and then searching in the partially-used interval you find.) But--- I don't see any way of making it faster to find games that have been created but not replied to. Random search doesn't seem promising either unless many more games are created than actually get played--- which may well be feasible.
I don't fully understand the mechanisms by which Freenet prevents duplicate keys, but I believe it is vulnerable to network partitioning (or just insufficient TTLs). In any case, heavy contention on a small portion of the namespace could quickly bring to light any flaws in the protocol or implementation.:)
A natural idea is to have an indexing service which keeps track of which games have been started but not replied to yet. But once you introduce such a mechanism, you might as well use that instead for game setup.
Dependence on perfect storage. The moves of a particular game will be scattered among a large collection of servers and duplicated. But the protocol as given can't handle the loss of _any_ records. There are some easy fixes, but this is almost a textbook example of the end-to-end principle: Freenet cannot (and does not) guarantee reliability, so the end-nodes need some mechanism for recovering from lost moves. The client can periodically check to see if the last move it made is still available, but this does allow for a subtle form of "cheating".
Games may also appear in retrospect to contain illegal moves if a move is lost by the system and and replaced--- but this is not a big problem, since the "cheating" can only be done by the player who would stand accused.
(I realize that the article was only meant as a sketch--- and that the author admits Freenet isn't reliable--- but the protocol can and should address how to handle the unreliability.)
Congestion control. Since Freenet includes no notification mechanisms (nor am I suggesting it should), clients must constantly check for keys they desire. This can naturally be tuned to provide some good compromise between response time and local resource usage. However, there is little way for a client to be notified that it is using too many resources elsewhere in the network. Freenet, just like the Internet, can suffer congestion collapse. (Rate-limiting implies dropping requests. Queueing only delays the time until you have to drop.)
Don't get me wrong--- I like Freenet; it has some very cool technical ideas. However, I think the claims of its usefulness (such as "uncensorable") are overstated. In particular, it's only effective against law-abiding censors, not totalitarian ones, who will cheerfully make running a Freenet node a punishable offense. Further, Freenet only changes the problem from "finding a copy of the forbidden content I want" to "finding the key for the forbidden content I want". (A smaller piece of information, certainly, but still one that has to be distributed outside of Freenet and is thus subject to interception and censorship.) Indexing is, in my opinion, a feature that any successful content-distribution system has to address.
You shouldn't believe everything the IPv6 people tell you. Sure, they _claim_ they will reduce the size of routing tables, but only by renumbering fairly often--- a scheme that has not been demonstrated on a large scale. (How often? Nobody knows.) Most of the recent growth in the size of routing tables has been from increased multihoming--- which IPv6 does not yet provide a good solution for.
IPv6 requires you to have a distinct range of IP addresses from each of your upstream ISPs. The addressing/routing architecture does not allow these ISPs to advertise your "other" prefix to their backbone providers (or, possibly, to their peers.) This negates much of the benefit of multihoming, since any particular address is tied to one ISPs--- and possibly to one ISP and one of that ISP's providers.
As far as I understand it, the current wisdom on IPv6 multihoming is to use tunnels between the various ISPs you have addresses for; this doesn't completely solve the problem, since you still have a dependence on the ISP which "owns" that particular address. And tunnelling, of course, adds extra overhead and an additional routing table entry in the ISP's routing tables.
IPv6 doesnt "solve" current problems with routing, it just attempts to legislate them out of existence. And yes, I _do_ subscribe to the IPv6-haters mailing list.
I'm reminded of the (sentient) ship Grey Area (?) in Iain M. Banks' "Excession" which turned itself into a museum of torture devices... The motivation was not explicitly spelled out, as I recall. But one interpretation was to justify its intervention in "primitive" cultures--- the Culture is generally nicer in dealing with bad people than the Grey Area was, and the ship was trying to argue that the sheer brutality of torture was sufficient cause for its actions.
There are lots of other sources of WWW latency other than packet switching latency. Given that most routers act at wire speed already, much of the delay on an HTTP request is already just round-trip-time, software processing, and queuing. It's unlikely that 25% of this latency occurs waiting for a switch to do its thing.
The last factor, queuing, is very important to switches, and is simply not addressed by this new technology. If you have two packets coming in that need to go out the same interface, one of them needs to be delayed; it's not clear how their switch handles this from the description in the article.
Still, it's nice to see optical networking technology that acts at a reasonable scale (nanoseconds rather than milliseconds or 100's of microseconds), but I'd like to see more details before I believe it delivers what it promises.
Google's advanced search page lets you search restricted to a particular site, which can be used to restrict to ".edu". Use their form, or do searches of the form "key word site:.edu"
I've experienced both sides of the question. Usually I can find anything I want on Google--- especially if it's technical information, but I've successfully looked up saints, theological arguments, gaming groups, etc. I occasionally supplement this with Citeseer, an excellent resource for research papers.
On the other hand, I was looking for a replacement rack mount kit for a Cisco switch that had been donated to my research group. Google and Altavista were pretty useless, as far as I could tell; I eventually just had to go to ECost and use their search facility to find the part I wanted.
So, I can see how users with different desires could easily develop widely divergent opinions about the utility of web search. Perhaps consumer sites are much less well searched? Perhaps one way that search engines can increase their utility is by making partnerships with online retailers to provide indexing of their product descriptions--- I'd be very happy if Amazon books or ECost electronics started showing up in response to my Google searches.
Unfortunately, not "struck down" yet. This was just the appeal on a preliminary injunction; the appeals court said the law can't be enforced yet. The court basically told the city what sort of evidence they would have to present in order to win their case... and that failing to do so, the lower court should issue a permanent injunction against the ordinance. So it's not quite dead yet--- there yet may be an actual trial.
Slashdot Mirror
The DMCA only covers copyrighted material, not material which is in the public domain.
This is not true. The anticircumvention portions talk about devices and technologies which are used to protect access to copyrighted works--- it is the technology itself which is illegal. It doesn't matter whether DeCSS is used to access copyrighted or non-copyrighted works, the "device" itself is still illegal. (The judge in the 2600 case said that this issue wasn't yet "ripe" because nobody's tried to prevent access to non-copyrighted works yet.)
Interestingly, a tool which is legal today (say, a DES cracker) might become illegal tomorrow if somebody starts using a content protection scheme which it defeats.
The author of the article isn't particularly careful about that distinction...
It's not clear that this _is_ theft of service, anyway (although many cable modem agreements state that you can't share the service with others.) Somebody is paying the bill for the pipe--- I don't believe that the ISP is owed money per device rather than for bandwdith. Would it still be "stealing" if a web proxy was used rather than NAT?
This only helps if the document is only in Google's cache. If it isn't, the attacker can just follow the link and read the original document.
Excluding entire documents that happen to match the checksum from the search results would work, but it'd be interesting to see how many false positives would result...
Not just the huddled masses. How many IP addresses do you have memorized? Even if you can still read slashdot, how much work will even the technically elite get done without email, which is very DNS-dependent? IP routing may go on, but my view is that DNS is a pretty central part of how the Internet works... Attacks on DNS are very scary because they interfere with the normal channels for responding to those attacks.
I'm working on a paper on this very topic (attacks against DNS infrastructure and how to defend against it). I don't think existing work on DDoS attacks is sufficient for the scale of assaults we may see in the future...
Please indicate where do you think that this breaks the RFCs.
RFC 1034, section 4.3.1 says that the legal responses (for a non-recursive query) are:
RRs that answer the question, together with an indication whether the data comes from a zone or is cached.
A referral to name servers which have zones which are closer ancestors to the name than the server sending the reply.
Even though giving back a "temporary error" is technically legal, it is not within the "spirit" of the specification to choose to do so always. Further, the algorithm in section 4.3.2 states:
Copy the NS RRs for the subzone into the authority section of the reply. Put whatever addresses are available into the additional section, using glue RRs if the addresses are not available from authoritative data or the cache. Go to step 4.
That is the idea. However, the RFC (2136) only requires that the changes be committed to stable storage--- and explicitly state that this may involve writing _just_ the changes, since writing out a huge text file could be expensive.
Horowitz has been writing columns for Salon since February 10, 1997--- hardly "recently". On the other hand, he is a loon who probably does little to change anybody's mind.
Heck, compared to the local San Francisco weeklies, Salon is the essence of evenhandedness. Do you have any online alternatives to suggest? I can't stand Slate (nor do I have any love for the company which owns it...)
Thanks for your coherent and insightful answer. I hadn't thought about Forth in quite that way, although it's been a long time since I played around with it. One could view unix piping as a concatenative programming language by defining "tokens" correctly...
Naturally, this behavior doesn't come without cost. With Forth (and Joy) it seems to me the cost of not having any arguments is that you can't do any type checking. Indeed, wouldn't any large Forth project have to document what "arguments" a word expects? I'm not convinced us poor humans can deal with all the power at our fingertips. :)
I think that one _could_ probably build a type-inference system into Joy. Basically all types would be parameterized--- rather than "int" it would be "stack of type X with int on top".
I like it, too. But, I think that the biggest problem is how close the Klingon homeworld is. If it's 4 days (!) away at warp 4, then it's just hours away with "modern" (TNG) warp technology.
This makes much of the future inconsistent. For example, why do no Klingon ships show up to battle the Borg? How come Earth never has anybody but the Enterprise to defend it from threat X--- isn't there any defensive fleet? (more of a problem with TOS.)
I'm also kinda annoyed that the doctor has to be a brand-new species--- could have used one of the throwaways from earlier shows. (Or did they?)
I think the benefits of Forth in terms of "thinking different" could just as easily be applied to other languages such as Perl, Scheme or ML. Each language has its own idioms and ways of approaching problems. Most of the points I've heard in favor of Forth boil down to "it's small", "it uses postfix", and "it forces you to write small functions." None of these seem particularly convincing to me.
I really can't understand what's so appealing about postfix notation as opposed to prefix or infix. (Despite what Chuck Moore says, most modern languages are actually a combination of the two, which does give Forth some edge in purity, especially for people who don't like parentheses.)
How would he then tell people how to look up the information on Freenet? If he already has an anonymous communication channel with which to distribute a Freenet key, then he could use that directly to distribute the information. If no such channel is available, then posting the key in a traceable manner will still expose him.
Someone could just try guessing random related names, of course, but I (or Intel) could just as easily guess what people would try and preemptively put information under those names as well...
I think there are enough users (in terms of Linux desktop installations) but not enough people who use exclusively non-Windows OSes. Anybody who dual-boots is going to be hard to sell to.
I gave up using Windows entirely (at home) about the time of my last hard drive crash, so I was naturally excited about Loki, and own two of their games. (And will probably buy Kohan if it's still available...) But I'm not a typical game-player--- and I can understand how someone who spends more time would have a Windows partition available and make use of it. Heck, I'm still tempted to buy Black and White and try to play it on my work laptop...
*sigh* I really hope I don't break my promise to give no more of my money to Bill G...
I've had @Home in the bay area for about 6 months, since moving to a new apartment meant I couldn't keep DSL. It's been an incredibly frustrating experience, with entire weeks in which we saw downtime for large periods of time every day. Our connectivity still has a tendency to migrate into a 70%-packet loss, 1000ms+ latency state and stay there for a while.
Not only that, but everyone in the apartment is frustrated by ATT @Home's tendency to treat us like idiots during every stage of our relationship with them, from the initial installation to complaints to customer service.
I'd switch back to DSL in a heartbeat (now that it's available in my area), except I really can't afford the extra cost to get PacBell's "Enhanced" service, nor do I want to fool around with the evil dynamic-IP PPP-over-Ethernet garbage they give with the Basic service these days.
Given that @Home needs more funding in order to stay afloat past the end of the year, according to an article yesterday in the San Jose Mercury Times, I wouldn't be surprised at all if their assets were acquired or if ATT switched partners.
The U.S. government is making itself look really bad with this incident, particularly at a time when it's trying to get U.S. scholars released from Chinese prisons. I'm surprised Russia and China haven't raised a stink about this, given that jurisdiction is a little shaky in this case--- has anybody seen articles about any reaction from the Russion government?
I'm very tempted to write my congresscritter and point out that arresting foreign citizens for "crimes" committed in their native land (where they are perfectly legal) is exactly the human rights violation we're protesting when China does it to U.S. citizens. Of course, the U.S. government has no problem being hypocritical anyway, but it still must look bad to the rest of the world...
However, these aren't the only parties--- I could easily see a judge dismissing the RIAA and SDMI from the case, but the complaint lists the U.S. Government (well, at least the AG) as targets as well. As others have pointed out, there are criminal penalties to be careful of, too. Given the recent arrest, I'd say there's a pretty "affirmative fear" there.
Does anyone know what the procedure would be if some of the "defendants" were dropped? Would the EFF have to refile or just amend their complaint?
(I don't like the newer Lego stuff with all the funny shaped pieces. I haven't decided whether it's because you really can't build anything other than the model they suggest with them, or whether I'm too old and dumb to come up with creative uses.)
What one should get from a CS degree is understanding, which necessarily involves almost every level of a system. Any physicist can learn to program, but most don't understand why, say, the order of nested loops matters when jumping through an array. Some theoreticians come up with elegant algorithms that are beaten by simpler ones which use a smaller working set. Machine language programmers sometimes lose sight of the fact that a better data structure would do more for performance than any pipeline optimization they can come up with. And C programmers can be baffled by the fact that their distributed system breaks despite the lack of obvious bugs on any single machine.
So, I agree that there are no "universal truths" to be discovered by assembly language programming. But you would go too far to say that there isn't any insight to be gained by being familiar enough with that level to do so--- after all, even in his new revision of the Art of Computer Programming, Knuth decided it was still worth expressing things in an assembly-like language to truly understand what the real performance costs were.
Now, you may have been trying to make the point that all these are really "software engineering" concerns, but I disagree that computer science is only the mathematical parts--- without the practical angle, CS is only so much hot air and might as well stay part of the math department.
A bigger problem is--- why does spinach break David, but being under water does not? Could David have been drinking all the water he wanted to?
I think this is a very clever design, and nicely illustrates how to use Freenet's signed-subspace keys to avoid interference from a third party. I do believe, however, that it suffers from a couple problems inherent in the design of Freenet--- although only the first looks like a showstopper to me.
Perhaps a better algorithm for searching could help (for example, finding the "next" number on game creation by doing a search for 1, 2, 4, 8, etc., and then searching in the partially-used interval you find.) But--- I don't see any way of making it faster to find games that have been created but not replied to. Random search doesn't seem promising either unless many more games are created than actually get played--- which may well be feasible.
I don't fully understand the mechanisms by which Freenet prevents duplicate keys, but I believe it is vulnerable to network partitioning (or just insufficient TTLs). In any case, heavy contention on a small portion of the namespace could quickly bring to light any flaws in the protocol or implementation. :)
A natural idea is to have an indexing service which keeps track of which games have been started but not replied to yet. But once you introduce such a mechanism, you might as well use that instead for game setup.
Games may also appear in retrospect to contain illegal moves if a move is lost by the system and and replaced--- but this is not a big problem, since the "cheating" can only be done by the player who would stand accused.
(I realize that the article was only meant as a sketch--- and that the author admits Freenet isn't reliable--- but the protocol can and should address how to handle the unreliability.)
- Congestion control. Since Freenet includes no notification mechanisms (nor am I suggesting it should), clients must constantly check for keys they desire. This can naturally be tuned to provide some good compromise between response time and local resource usage. However, there is little way for a client to be notified that it is using too many resources elsewhere in the network. Freenet, just like the Internet, can suffer congestion collapse. (Rate-limiting implies dropping requests. Queueing only delays the time until you have to drop.)
Don't get me wrong--- I like Freenet; it has some very cool technical ideas. However, I think the claims of its usefulness (such as "uncensorable") are overstated. In particular, it's only effective against law-abiding censors, not totalitarian ones, who will cheerfully make running a Freenet node a punishable offense. Further, Freenet only changes the problem from "finding a copy of the forbidden content I want" to "finding the key for the forbidden content I want". (A smaller piece of information, certainly, but still one that has to be distributed outside of Freenet and is thus subject to interception and censorship.) Indexing is, in my opinion, a feature that any successful content-distribution system has to address.You shouldn't believe everything the IPv6 people tell you. Sure, they _claim_ they will reduce the size of routing tables, but only by renumbering fairly often--- a scheme that has not been demonstrated on a large scale. (How often? Nobody knows.) Most of the recent growth in the size of routing tables has been from increased multihoming--- which IPv6 does not yet provide a good solution for.
IPv6 requires you to have a distinct range of IP addresses from each of your upstream ISPs. The addressing/routing architecture does not allow these ISPs to advertise your "other" prefix to their backbone providers (or, possibly, to their peers.) This negates much of the benefit of multihoming, since any particular address is tied to one ISPs--- and possibly to one ISP and one of that ISP's providers.
As far as I understand it, the current wisdom on IPv6 multihoming is to use tunnels between the various ISPs you have addresses for; this doesn't completely solve the problem, since you still have a dependence on the ISP which "owns" that particular address. And tunnelling, of course, adds extra overhead and an additional routing table entry in the ISP's routing tables.
IPv6 doesnt "solve" current problems with routing, it just attempts to legislate them out of existence. And yes, I _do_ subscribe to the IPv6-haters mailing list.
I'm reminded of the (sentient) ship Grey Area (?) in Iain M. Banks' "Excession" which turned itself into a museum of torture devices... The motivation was not explicitly spelled out, as I recall. But one interpretation was to justify its intervention in "primitive" cultures--- the Culture is generally nicer in dealing with bad people than the Grey Area was, and the ship was trying to argue that the sheer brutality of torture was sufficient cause for its actions.
There are lots of other sources of WWW latency other than packet switching latency. Given that most routers act at wire speed already, much of the delay on an HTTP request is already just round-trip-time, software processing, and queuing. It's unlikely that 25% of this latency occurs waiting for a switch to do its thing.
The last factor, queuing, is very important to switches, and is simply not addressed by this new technology. If you have two packets coming in that need to go out the same interface, one of them needs to be delayed; it's not clear how their switch handles this from the description in the article.
Still, it's nice to see optical networking technology that acts at a reasonable scale (nanoseconds rather than milliseconds or 100's of microseconds), but I'd like to see more details before I believe it delivers what it promises.
Google's advanced search page lets you search restricted to a particular site, which can be used to restrict to ".edu". Use their form, or do searches of the form "key word site:.edu"
I've experienced both sides of the question. Usually I can find anything I want on Google--- especially if it's technical information, but I've successfully looked up saints, theological arguments, gaming groups, etc. I occasionally supplement this with Citeseer, an excellent resource for research papers.
On the other hand, I was looking for a replacement rack mount kit for a Cisco switch that had been donated to my research group. Google and Altavista were pretty useless, as far as I could tell; I eventually just had to go to ECost and use their search facility to find the part I wanted.
So, I can see how users with different desires could easily develop widely divergent opinions about the utility of web search. Perhaps consumer sites are much less well searched? Perhaps one way that search engines can increase their utility is by making partnerships with online retailers to provide indexing of their product descriptions--- I'd be very happy if Amazon books or ECost electronics started showing up in response to my Google searches.
Unfortunately, not "struck down" yet. This was just the appeal on a preliminary injunction; the appeals court said the law can't be enforced yet. The court basically told the city what sort of evidence they would have to present in order to win their case... and that failing to do so, the lower court should issue a permanent injunction against the ordinance. So it's not quite dead yet--- there yet may be an actual trial.