I've played chess very occasionally on freechess.org. If you monitor the main forum, you'll see bots occasionally announcing that they will be starting a lecture in a few minutes, whether it be on endgame technique or a particular notable historical game. The ones I've seen seem to last in the neighborhood of an hour, though I've never actually "attended" any.
I was unaware that was the behavior. I guess that means Slashdot moderation is reactionary -- it takes into account both the article and the current score. (Come to think of it, this is how I moderate on Slashdot.:-) If the score is ok, then you don't register agreement with that by modding it some more.
Which is the opposite of how Kuro5hin works; you just vote on each comment what you think it deserves.
The difference being that Slashdot mod points are a precious resource; Kuro5hin mod points are as free as dirt.
Maybe this spells doom for averaging on Slashdot.:-) I'm still curious as to why CT and Hemos are so opposed to it from a performance standpoint, when performance is not really even a consideration.
I'm curious to see an example demonstrating your claim. I can't come up with any situation off the top of my head where summation is provably better at countering groupthink than averaging.
In K5, everyone can indeed rate posts, which, as you point out, could be detrimental to the rating scheme -- especially in a forum as large as Slashdot.
I'm happy keeping Slashdot's moderation/metamoderation scheme as-is (moderators chosen on basis of karma, metamoderation keeping them in check). I'm not suggesting any change along those lines. Averaging comments is a distinct idea, and it turns out to be more significant than just "different math".
Gedanken experiment:
Consider a comment which receives the following ratings:
+1: 10 votes -1: 4 votes
As I currently understand Slashcode to work (caveat: I have never looked at the code!), Slashdot would rate the comment as +5, which is unrealistic given the negative response the comment elicited. Averaging would rate the comment as +3, which is much more reasonable. Er, given the assumption that all moderations are in good faith (an assumption upon which moderation seems to depend).
I, too, rarely make use of IRC and will not be online tonight. . ..
5) During the infamous CT/S11 IRC discussion, the topic of scoring behavior was brought up. Specifically, the difference between Slashdot (which sums ratings to compute a comment's score) and Kuro5hin (which averages ratings to compute a comment's score) was pointed out.
In my opinion, this is a very crucial difference between the two systems. I also believe that the Kuro5hin system is much more accurate at capturing readers' ratings of a comment.
Both CommanderTaco and Hemos emphasized strongly that the load of computing averages would break their server. I am confused at this; there is a simple way of updating averages. The current scheme looks like this:
score = score + diff
In order to update an average, it can be computed at the time of the moderation action using the following logic:
(Note that "score" must be stored in floating-point, though it may certainly be rounded for display.) The math at the time of moderation will indeed require a few extra instructions, but this is insignificant; the concern of overhead is at comment display time, which this scheme entirely avoids. Yes, there will be slight rounding error, which should nevertheless not prove significant in the context of moderation.
My question for CT and Hemos is this: Why (in light of this proposed code) not compute scores using averaging? I see no compelling reason that summation should be used. My personal opinion is that this change (along with hiding karma;-) would improve comment quality a great deal.
Granted. In fact, I'm sure they'd be happy in practice to see it, would gladly give permission, etc. In principle, though, it is a violation of their copyright to do so without permission. I just found it a very ironic juxtaposition, especially as the comment was highly modded.
I find it ironic that Mr. Goldstein points out in this article that his magazine is copyrighted. Here we have a verbatim copy of it, without explicit attribution. Moreover, do you think that the average Slashdot reader is too lazy or imbecile to notice and click on the article link in the story? This comment deserves a -1: Redundant at very least. Hairy_Potter is karma whoring.
You seem to be asking one question and giving a specific example of another.
The question you ask is what to do when our ideals conflict with the real world. In that case, if you value your ideals highly enough, you quit your job. Consider the ACM Code of Ethics as an example. If you cannot uphold its ideals, you should, ethically, discontinue your job. The same goes here. If you cannot consionably do this (i.e., the very concept of a filter is wrong in your sight) then don't do the job!
The question you seem to be asking by your example is how to design this filter so that it doesn't violate your standards of censorship (i.e., porn is not allowed, but anything else, in general, is). I am not well equipped to answer this question, though I would suggest you consider designs that function on the basis of individual site / page rejection, rather than some sort of content heuristic. In general, the former seem to be much better than the latter, though they depend on a great deal of input into determining excluded sites.
There existed at least one such service, CrossingGuard, which seems now to have been discontinued. I cannot speak to its effectiveness, however, so I'm willing to be shouted down on this one point.:-)
Billing purposes: you know exactly what application (er, copy of Linux) ran for such and such an amount of time.
Transaction-based environments: With applications popping their heads up and down to handle brief and frequent transactions, it's better to let MVS take care of the headaches of swapping Linux images in and out of memory than to use Linux to handle that.
Disadvantages: If all you need is brute processing power (i.e., you aren't doing any transaction-oriented stuff), then run a single copy; you'll get better mileage.
What's so cool about this is that IBM is charging a flat rate for customers to install this on their systems, regardless of how much it is used. Previously, all IBM operating software has been billed on a monthly basis (essentially, you "rent" the operating system), on a pricing scale depending on how many processors you had, how many virtual machine partitions you had, etc.
11223 is trolling... The advantages of mainrames in terms of brute processing power, reliability, etc., are well known. I'm not claiming categorically that they outstrip PC's; there are applications much better run on PC's. There are, however, many applications for which a mainframe is the only sensible choice.
I work with mainframes every day. They aren't exactly the monstrous monolithic beasts that we picture from yesteryear. They can be pretty darn nimble. And wherever transactions, reliability, paralleleism, or sheer number of simultaneous connections is needed, they kick any PC's butt.
Even if you could magically mutate all of your cells' genes at once to match some new yardstick (e.g., increased endurance), that would not have an immediate physiological change. Take a simpler example: change your genes to give you blue eyes instead of mud-brown. Your eyes stay mud-brown, since the pigment produced by your cells doesn't suddenly and magically chane to blue. (I don't know how the pigment works, so I won't comment on how quickly your eyes would become blue.)
The same goes here. To acquire resistance to a certain disease, or build stronger bone, or any such thing: even if the gene shift is instantaneous, the effects will be delayed.
At the time I couldn't recall the one specific project I was thinking of, but a brief search dredged it up... AT&T's Crowds, a really nifty idea.
I had two mis-recollections from senior sem presentations I saw last year... I thought Anonymizer, wasn't a commercial project, but it turns out that it is. I'm not familiar with them at all, so I can't speak to their approach. I'd also thought that onion-routing was a little beyond the experimental stage, but I guess I was wrong.
Indeed... Little specifics are given as to the details of the protocols, routing methods, etc. IMHO, the routing techniques he describes will incur significant overhead. I don't think this service would be useful to many people.
Better to check out existing services that provide anonymity, security, etc. There's a bunch out there that are already much better established and much better thought-through.
Afterthought: is it really a good idea to implement these ideas on the protocol level? I would think that abstraction principles would dictate that security and anonymity are better implemented in higher levels of abstraction...
The key word is "marketplace". Since GPL'd software is made available without cost, I wouldn't think this could be made to apply to GPL'd software. The notion is that by selling something, you give up rights. But by not selling it, presumably you retain them.
I think you're over-simplifying gravity. Yes, two-body gravitational systems are deterministic (but then, in the absence of a frame of reference, they're quite trivial, too). But even the three-body problem is non-deterministic. Moreover, the proposed scenario considers gravity on a scale much bigger than even galactic superclusters. Picture such a supercluster as perhaps a single "cell" in a massive organism, or maybe not even more than a complex "protein" or somesuch. The long-ranging effects of gravity over long periods of time then become quite complex and interesting; certainly more than enough to provide a level of complexity that we would associate with an intelligent agent.
Consider, too, the nature of our world. Nearly every interaction we experience can be described completely in terms of the gravitational and electromagnetic forces. Since gravity is an aggregate force with negligible impact on the particle level, our atoms (stars? galaxies?) live their life completely in the domain of electromagnetism. Atomic electrical forces are not appreciably more complex than interplanetary gravity.
I think that gravitational intelligences are actually quite a plausible idea except, as some have mentioned already, that the long-term fate of the universe severely constrains their development.
I would argue that Jon is not stereotyping so much as he is summarizing statistics. While it is true that there is no single model for any particular type of internet-user (be they male, female, "geek", or middle-class, etc.), there are overall trends. For the purposes of social analysis, Jon is summarizing the most common members of a particular category. I don't feel that it is entirely necessary for him to qualify... "This does not necessarily apply to those women who: are not members of academia, live in Argentina, do not make more than 100k a year, do not make less than 10k a year, are married with 4 or more children, etc. etc."
I do wish, however, that there were a lot less arm-waving in what he says. While it makes sense, there are no real statistics to back it up. We are left guessing that his intent was to portray a statistical social analysis, rather than knowing exactly what he intended to convey, quantitatively.
I sincerely hope this is successful, too. However, if there are people out there who want it (and from the attitudes expressed by NC leaders, there certainly are), there are people out there who will build it. Even if Pinkerton listens to the voices of reason and stops this scary venture, someone will step up in their place to serve the needs of their country . . . cough . . . make some money.
I'm overjoyed that Pinkerton is at least being open to listening to the public. Would that everyone were so responsive. Some day, however, something like this will happen and the only way to fight it will be through legal routes. This project is in violation of constitutional rights, and I hope and pray that when the day comes there will be people willing to fight it, lawyers willing to take the task on, and jurists perceptive of the gravity of this.
At first blush this sounds like a really cool idea!
Until a new form of monetary exchange is invented however, this isn't really feasible. Once his credit charge is converted into cash (or some non-corporeal form thereof) there is no way of enforcing his restrictions.
There's another problem; can you imagine billions upon billions of dollars simply disappearing into the unknown? This would happen if the collective restrictions of each person who uses them along the way pile up... Eventually there would be no permissible use for the money. Example: Bob gives Sally a dollar that must be spent on research and development. Sally spends it at lsl.com buying the latest copy of her favorite Linux distribution, but forces them to spend it on food. Presto! The intersection of R&D and food is null. The dollar has vaporized.
People might also just set up clearinghouses to exchange restricted dollars for non-restricted ones. You have a "food" dollar, Kim needs food. She gives you her "computer hardware" dollar in exchange for your "food" dollar.
Moreover, this actually makes the value of money mutable. Certainly a "food" dollar is worth more than a "spinach" dollar! Imagine the confusion this would breed!
I like the idea, though... Perhaps there are more practical applications for it...
30 qubits would not be nearly enough to factor a large number (whether prime or not;-). Quantum computers store their entire state in the quantum register. This includes not merely data that is currently being operated on (as with traditional computer registers), but also all other state information. The current "instruction" or step is coded in the quantum register, as are all "variables" (including ones that no longer need use; quantum theory prevents us from doing nifty things like "let a = b", which actually makes it harder to reuse register space).
It's more meaningful, therefore, to think of the number of qubits in a quantum computer as its memory size, not its addressing capacity. A 30-qubit quantum computer isn't nearly analagous to a 32-bit von Neumann machine; it's much closer to one with 4 *bytes* of RAM!
In reality, it's been estimated that before we can factor reasonably large numbers, we'll need quantum machines possessing hundreds of bits. Quantum programmers of tomorrow will get their kick out of squeezing as much as possible out of their half-kilobyte of register space.
Random observation: we're seeing e-everything nowadays. Tomorrow will we start seeing q-everything?
I think what is being overlooked here is the nature of books as a communications medium. Books are a means of communication from the author to his or her audience. They are inherently one-way communication. The push for interactivity in any communications medium overturns this model and attempts to establish two-way communication.
In many ways, much of this two-way communication is no better than a choose-your-own-ending novel. We, the readers, aren't actually communicating back with the author; we're really just communicating with an agent the author has created, whether it be a simple choice of storyline thread, or something more interactive like the classic Adventure game. (Caveat: perhaps I'm not understanding what Jon means by interactivity.)
I don't believe that books should be retrofitted with an interactive model. We love books, magazines, and newspapers precisely because they provide this one-way communication. While a "choose-your-own-ending" Tom Clancey novel might be fun, there will always be a greater market for the non-interactive variety.
Storytelling is important. Granted, it's not given as much importance by Western culture as in other cultures, but it's still a very central part of our lives. I think that moving books to the web will in the end change very little about the way in which stories are told -- not because we're merely accustomed to it but because it's how we're wired. Though the delivery medium has changed much over the past centuries, stories have always been -- and always will be -- loved for their narrative form.
I believe that more interactive forms of communication will be important as the new media continue to metamorphose. However, I think these will grow parallel to the digitization of books, not at their expense. Attempting to retrofit two-way communication onto books may work, but in the end will be unsuccessful; completely new means must be developed for this purpose.
According to the article they have no legal basis for doing anything.... There are actually different contracts for CA and OR, which contain no language that compels people to repay upon cancellation. IANAL, but I don't see how they could build a case.
Slashdot Mirror
I've played chess very occasionally on freechess.org. If you monitor the main forum, you'll see bots occasionally announcing that they will be starting a lecture in a few minutes, whether it be on endgame technique or a particular notable historical game. The ones I've seen seem to last in the neighborhood of an hour, though I've never actually "attended" any.
I call your bluff; evidently you do not work for the same IBM that I do.
No IBM internal servers have been blocked, nor access to external servers.
I was unaware that was the behavior. I guess that means Slashdot moderation is reactionary -- it takes into account both the article and the current score. (Come to think of it, this is how I moderate on Slashdot. :-) If the score is ok, then you don't register agreement with that by modding it some more.
Which is the opposite of how Kuro5hin works; you just vote on each comment what you think it deserves.
The difference being that Slashdot mod points are a precious resource; Kuro5hin mod points are as free as dirt.
Maybe this spells doom for averaging on Slashdot. :-) I'm still curious as to why CT and Hemos are so opposed to it from a performance standpoint, when performance is not really even a consideration.
I'm curious to see an example demonstrating your claim. I can't come up with any situation off the top of my head where summation is provably better at countering groupthink than averaging.
Or, maybe I've just been trolled. :-)
In K5, everyone can indeed rate posts, which, as you point out, could be detrimental to the rating scheme -- especially in a forum as large as Slashdot.
I'm happy keeping Slashdot's moderation/metamoderation scheme as-is (moderators chosen on basis of karma, metamoderation keeping them in check). I'm not suggesting any change along those lines. Averaging comments is a distinct idea, and it turns out to be more significant than just "different math".
Gedanken experiment:
Consider a comment which receives the following ratings:
As I currently understand Slashcode to work (caveat: I have never looked at the code!), Slashdot would rate the comment as +5, which is unrealistic given the negative response the comment elicited. Averaging would rate the comment as +3, which is much more reasonable. Er, given the assumption that all moderations are in good faith (an assumption upon which moderation seems to depend).
I, too, rarely make use of IRC and will not be online tonight. . . .
5) During the infamous CT/S11 IRC discussion, the topic of scoring behavior was brought up. Specifically, the difference between Slashdot (which sums ratings to compute a comment's score) and Kuro5hin (which averages ratings to compute a comment's score) was pointed out.
In my opinion, this is a very crucial difference between the two systems. I also believe that the Kuro5hin system is much more accurate at capturing readers' ratings of a comment.
Both CommanderTaco and Hemos emphasized strongly that the load of computing averages would break their server. I am confused at this; there is a simple way of updating averages. The current scheme looks like this:
In order to update an average, it can be computed at the time of the moderation action using the following logic:
(Note that "score" must be stored in floating-point, though it may certainly be rounded for display.) The math at the time of moderation will indeed require a few extra instructions, but this is insignificant; the concern of overhead is at comment display time, which this scheme entirely avoids. Yes, there will be slight rounding error, which should nevertheless not prove significant in the context of moderation.
My question for CT and Hemos is this: Why (in light of this proposed code) not compute scores using averaging? I see no compelling reason that summation should be used. My personal opinion is that this change (along with hiding karma ;-) would improve comment quality a great deal.
Granted. In fact, I'm sure they'd be happy in practice to see it, would gladly give permission, etc. In principle, though, it is a violation of their copyright to do so without permission. I just found it a very ironic juxtaposition, especially as the comment was highly modded.
I find it ironic that Mr. Goldstein points out in this article that his magazine is copyrighted. Here we have a verbatim copy of it, without explicit attribution. Moreover, do you think that the average Slashdot reader is too lazy or imbecile to notice and click on the article link in the story? This comment deserves a -1: Redundant at very least. Hairy_Potter is karma whoring.
You seem to be asking one question and giving a specific example of another.
The question you ask is what to do when our ideals conflict with the real world. In that case, if you value your ideals highly enough, you quit your job. Consider the ACM Code of Ethics as an example. If you cannot uphold its ideals, you should, ethically, discontinue your job. The same goes here. If you cannot consionably do this (i.e., the very concept of a filter is wrong in your sight) then don't do the job!
The question you seem to be asking by your example is how to design this filter so that it doesn't violate your standards of censorship (i.e., porn is not allowed, but anything else, in general, is). I am not well equipped to answer this question, though I would suggest you consider designs that function on the basis of individual site / page rejection, rather than some sort of content heuristic. In general, the former seem to be much better than the latter, though they depend on a great deal of input into determining excluded sites.
There existed at least one such service, CrossingGuard, which seems now to have been discontinued. I cannot speak to its effectiveness, however, so I'm willing to be shouted down on this one point. :-)
Advantages:
Disadvantages: If all you need is brute processing power (i.e., you aren't doing any transaction-oriented stuff), then run a single copy; you'll get better mileage.
What's so cool about this is that IBM is charging a flat rate for customers to install this on their systems, regardless of how much it is used. Previously, all IBM operating software has been billed on a monthly basis (essentially, you "rent" the operating system), on a pricing scale depending on how many processors you had, how many virtual machine partitions you had, etc.
11223 is trolling... The advantages of mainrames in terms of brute processing power, reliability, etc., are well known. I'm not claiming categorically that they outstrip PC's; there are applications much better run on PC's. There are, however, many applications for which a mainframe is the only sensible choice.
I work with mainframes every day. They aren't exactly the monstrous monolithic beasts that we picture from yesteryear. They can be pretty darn nimble. And wherever transactions, reliability, paralleleism, or sheer number of simultaneous connections is needed, they kick any PC's butt.
Apache-on-a-palm was posted to the Apache section, too.
Even if you could magically mutate all of your cells' genes at once to match some new yardstick (e.g., increased endurance), that would not have an immediate physiological change. Take a simpler example: change your genes to give you blue eyes instead of mud-brown. Your eyes stay mud-brown, since the pigment produced by your cells doesn't suddenly and magically chane to blue. (I don't know how the pigment works, so I won't comment on how quickly your eyes would become blue.)
The same goes here. To acquire resistance to a certain disease, or build stronger bone, or any such thing: even if the gene shift is instantaneous, the effects will be delayed.
At the time I couldn't recall the one specific project I was thinking of, but a brief search dredged it up... AT&T's Crowds, a really nifty idea.
I had two mis-recollections from senior sem presentations I saw last year... I thought Anonymizer, wasn't a commercial project, but it turns out that it is. I'm not familiar with them at all, so I can't speak to their approach. I'd also thought that onion-routing was a little beyond the experimental stage, but I guess I was wrong.
Indeed... Little specifics are given as to the details of the protocols, routing methods, etc. IMHO, the routing techniques he describes will incur significant overhead. I don't think this service would be useful to many people.
Better to check out existing services that provide anonymity, security, etc. There's a bunch out there that are already much better established and much better thought-through.
Afterthought: is it really a good idea to implement these ideas on the protocol level? I would think that abstraction principles would dictate that security and anonymity are better implemented in higher levels of abstraction...
The key word is "marketplace". Since GPL'd software is made available without cost, I wouldn't think this could be made to apply to GPL'd software. The notion is that by selling something, you give up rights. But by not selling it, presumably you retain them.
I think you're over-simplifying gravity. Yes, two-body gravitational systems are deterministic (but then, in the absence of a frame of reference, they're quite trivial, too). But even the three-body problem is non-deterministic. Moreover, the proposed scenario considers gravity on a scale much bigger than even galactic superclusters. Picture such a supercluster as perhaps a single "cell" in a massive organism, or maybe not even more than a complex "protein" or somesuch. The long-ranging effects of gravity over long periods of time then become quite complex and interesting; certainly more than enough to provide a level of complexity that we would associate with an intelligent agent.
Consider, too, the nature of our world. Nearly every interaction we experience can be described completely in terms of the gravitational and electromagnetic forces. Since gravity is an aggregate force with negligible impact on the particle level, our atoms (stars? galaxies?) live their life completely in the domain of electromagnetism. Atomic electrical forces are not appreciably more complex than interplanetary gravity.
I think that gravitational intelligences are actually quite a plausible idea except, as some have mentioned already, that the long-term fate of the universe severely constrains their development.
I would argue that Jon is not stereotyping so much as he is summarizing statistics. While it is true that there is no single model for any particular type of internet-user (be they male, female, "geek", or middle-class, etc.), there are overall trends. For the purposes of social analysis, Jon is summarizing the most common members of a particular category. I don't feel that it is entirely necessary for him to qualify... "This does not necessarily apply to those women who: are not members of academia, live in Argentina, do not make more than 100k a year, do not make less than 10k a year, are married with 4 or more children, etc. etc."
I do wish, however, that there were a lot less arm-waving in what he says. While it makes sense, there are no real statistics to back it up. We are left guessing that his intent was to portray a statistical social analysis, rather than knowing exactly what he intended to convey, quantitatively.
I sincerely hope this is successful, too. However, if there are people out there who want it (and from the attitudes expressed by NC leaders, there certainly are), there are people out there who will build it. Even if Pinkerton listens to the voices of reason and stops this scary venture, someone will step up in their place to serve the needs of their country . . . cough . . . make some money.
I'm overjoyed that Pinkerton is at least being open to listening to the public. Would that everyone were so responsive. Some day, however, something like this will happen and the only way to fight it will be through legal routes. This project is in violation of constitutional rights, and I hope and pray that when the day comes there will be people willing to fight it, lawyers willing to take the task on, and jurists perceptive of the gravity of this.
At first blush this sounds like a really cool idea!
Until a new form of monetary exchange is invented however, this isn't really feasible. Once his credit charge is converted into cash (or some non-corporeal form thereof) there is no way of enforcing his restrictions.
There's another problem; can you imagine billions upon billions of dollars simply disappearing into the unknown? This would happen if the collective restrictions of each person who uses them along the way pile up... Eventually there would be no permissible use for the money. Example: Bob gives Sally a dollar that must be spent on research and development. Sally spends it at lsl.com buying the latest copy of her favorite Linux distribution, but forces them to spend it on food. Presto! The intersection of R&D and food is null. The dollar has vaporized.
People might also just set up clearinghouses to exchange restricted dollars for non-restricted ones. You have a "food" dollar, Kim needs food. She gives you her "computer hardware" dollar in exchange for your "food" dollar.
Moreover, this actually makes the value of money mutable. Certainly a "food" dollar is worth more than a "spinach" dollar! Imagine the confusion this would breed!
I like the idea, though... Perhaps there are more practical applications for it...
30 qubits would not be nearly enough to factor a large number (whether prime or not ;-). Quantum computers store their entire state in the quantum register. This includes not merely data that is currently being operated on (as with traditional computer registers), but also all other state information. The current "instruction" or step is coded in the quantum register, as are all "variables" (including ones that no longer need use; quantum theory prevents us from doing nifty things like "let a = b", which actually makes it harder to reuse register space).
It's more meaningful, therefore, to think of the number of qubits in a quantum computer as its memory size, not its addressing capacity. A 30-qubit quantum computer isn't nearly analagous to a 32-bit von Neumann machine; it's much closer to one with 4 *bytes* of RAM!
In reality, it's been estimated that before we can factor reasonably large numbers, we'll need quantum machines possessing hundreds of bits. Quantum programmers of tomorrow will get their kick out of squeezing as much as possible out of their half-kilobyte of register space.
Random observation: we're seeing e-everything nowadays. Tomorrow will we start seeing q-everything?
I think what is being overlooked here is the nature of books as a communications medium. Books are a means of communication from the author to his or her audience. They are inherently one-way communication. The push for interactivity in any communications medium overturns this model and attempts to establish two-way communication.
In many ways, much of this two-way communication is no better than a choose-your-own-ending novel. We, the readers, aren't actually communicating back with the author; we're really just communicating with an agent the author has created, whether it be a simple choice of storyline thread, or something more interactive like the classic Adventure game. (Caveat: perhaps I'm not understanding what Jon means by interactivity.)
I don't believe that books should be retrofitted with an interactive model. We love books, magazines, and newspapers precisely because they provide this one-way communication. While a "choose-your-own-ending" Tom Clancey novel might be fun, there will always be a greater market for the non-interactive variety.
Storytelling is important. Granted, it's not given as much importance by Western culture as in other cultures, but it's still a very central part of our lives. I think that moving books to the web will in the end change very little about the way in which stories are told -- not because we're merely accustomed to it but because it's how we're wired. Though the delivery medium has changed much over the past centuries, stories have always been -- and always will be -- loved for their narrative form.
I believe that more interactive forms of communication will be important as the new media continue to metamorphose. However, I think these will grow parallel to the digitization of books, not at their expense. Attempting to retrofit two-way communication onto books may work, but in the end will be unsuccessful; completely new means must be developed for this purpose.
According to the article they have no legal basis for doing anything.... There are actually different contracts for CA and OR, which contain no language that compels people to repay upon cancellation. IANAL, but I don't see how they could build a case.
Alas, the trouble with spam is that it is such a cheap tool that it doesn't amount to much of a deduction / penalty anyway.