Bah! The ArXiv serves a completely different purpose from that of a peer-reviewed journal. *Anything* can (and does) appear on the ArXiv; this is a feature and a serious shortcoming.
There is a reason that peer-review journals will remain the gold standard for acadmic research: when it appears in such a place we have reason to belive that it is correct!
This is a topic dear to my heart, so I thought I would air a couple of comments.
First of all, the actual economics of the journal publication business are a bit different from what is suggested in the original post: very rare is the author that pays page fees out of pocket. The unwritten rule is that if an author can pay the page fees from a grant (of if the author's university has a policy of defraying these costs), they are payed; otherwise they are simply not paid. (It should be clear that it would be a self destructive strategy for journals to actually require assistant professors to ante up $5000 for a 20 page journal article!)
The current system is, however, clearly outdated and patently unfair: the academic community has allowed the journal publishing business to apply (copyright, pricing) practices, born in the era of paper and print publication, to the current age, marked in this context by the fact that distribution can be effected free-of-charge. The existing model is one in which, roughly, authors and reviewers, who are doing all the hard technical work that make the journal valuable, are payed nothing and editors (also academics), who maintain the journal's quality standards, are payed a nominal annual stipend. You would be right, then, to wonder where all the money goes that is garnered from a, say, $1700 annual library subscription fee. This goes to the publisher (whose costs are not zero, incidentally). Knuth's letter, written at the time that the Journal of Algorithms board resigned in response to irresponsible pricing on the part of Elsevier, is a good read on this topic (it is linked to from the TOC website below).
One natural response to this is for academic and professional societies to take up the task which, in the case of CS, has happened with great success (e.g., ACM/SIAM/IEEE). They have adopted the pricing strategy above, either maintaining lower subscription costs or passing the profits along to a good cause (the society).
A more dramatic response is that taken by a new CS journal (Theory of Computing): (i.) maintain zero cash flow, (ii.) adopt the internet as the primary means of dissemination and (even more radically) (iii.) leave copyright with the authors. See also the Electronic Journal of Combinatorics.
It is interesting to note that in order for this to be possible, not only does dissemination have to be free, but authors must typeset their own articles. If you are a member of the math/cs/physics community, you have been doing this yourself for perhaps 20 years.
Let us hope that a rapid cultural evolution divorces the publishing houses of these funds that the academic community can better spend on research and education!
First of all, I have to say that if history is any guide, this proof is wrong. Incorrect claimed proofs of the twin prime conjecture outnumber correct proofs by a rather large margin:).
To calibrate this claim a bit it is worth noting that, to date, no one has even been able to prove that there are infinitely many primes pairs within distance o(log n); it has actually been proven for a constant c less than 1, there are finitely many pairs withing distance c log n, which I think is quite dramatic---I don't remember if this requires the Riemann Hypothesis or not.
(By the way, one say that there are infinitely many prime pairs withing distance f(n) of each other if there are infinitely many pairs (a_i, b_i) (both prime) so that |a_i - b_i | max(f(a_i),f(b_i)).)
Browsing through the above comments on this topic leaves me thoroughly discouraged with the Slashdot readership. SCO is *obviously* to blame; they probably thought the locks constituted IP infringement and arranged to pull the plug.
Presumably the marketing/bottom-line folks at RH have carefully considered the decision to cease boxed set sales. There *is* something romantic about the boxed sets, but I can understand their decision.
What does seem surprising (suspicious?) is the decision to leave package maintenance to the "developers themselves," without a period when the packages are under RH's control. This sounds a bit untenable, considering that RH's major value added feature is a distribution in which the (many) various components have undergone at least some kind of testing. I can't help but wonder how they will manage this.
Of course, opening up the development lists to the public seems like a positive move.
I guess we'll just have to wait until tomorrow to see what the details are...
The most puzzling aspect of this story is that the job of mapping the US internet is sufficient to earn a Ph.D. Of course, it is possible that there are aspects of the author's thesis that go beyond what is advertised above.
I admit that this author is not alone--in the CS department where I work, "experimental" Ph.D. theses featuring poorly designed experiments or no scientific work at all (which appears to be the case above) are a constant problem.
Perhaps this is an accident of the youth of the field.
What is being advertised here is not unbreakable in the sense used by most mathematician or serious cryptographers. (When a cryptographer says unbreakable, s/he means that the system is secure even against an adversary with unlimited computing power.)
Ideal use of a one time pad does have this property. There was a nice breakthrough in the EuroCrypt conference last year, where it was shown that one can obtain similar behavior even with keys that are shorter than the message to be encrypted, as long as the messages that you wish to encrypt are fairly random.
In any case, if you'd like to really understand what is going on here, for goodness' sake don't bother with Schneier's book; have a look at Goldreich's, "Foundations of Cryptography".
Slashdot Mirror
Bah! The ArXiv serves a completely different purpose from that of a peer-reviewed journal. *Anything* can (and does) appear on the ArXiv; this is a feature and a serious shortcoming.
There is a reason that peer-review journals will remain the gold standard for acadmic research: when it appears in such a place we have reason to belive that it is correct!
This is a topic dear to my heart, so I thought I would air a couple of comments.
First of all, the actual economics of the journal publication business are a bit different from what is suggested in the original post: very rare is the author that pays page fees out of pocket. The unwritten rule is that if an author can pay the page fees from a grant (of if the author's university has a policy of defraying these costs), they are payed; otherwise they are simply not paid. (It should be clear that it would be a self destructive strategy for journals to actually require assistant professors to ante up $5000 for a 20 page journal article!)
The current system is, however, clearly outdated and patently unfair: the academic community has allowed the journal publishing business to apply (copyright, pricing) practices, born in the era of paper and print publication, to the current age, marked in this context by the fact that distribution can be effected free-of-charge. The existing model is one in which, roughly, authors and reviewers, who are doing all the hard technical work that make the journal valuable, are payed nothing and editors (also academics), who maintain the journal's quality standards, are payed a nominal annual stipend. You would be right, then, to wonder where all the money goes that is garnered from a, say, $1700 annual library subscription fee. This goes to the publisher (whose costs are not zero, incidentally). Knuth's letter, written at the time that the Journal of Algorithms board resigned in response to irresponsible pricing on the part of Elsevier, is a good read on this topic (it is linked to from the TOC website below).
One natural response to this is for academic and professional societies to take up the task which, in the case of CS, has happened with great success (e.g., ACM/SIAM/IEEE). They have adopted the pricing strategy above, either maintaining lower subscription costs or passing the profits along to a good cause (the society).
A more dramatic response is that taken by a new CS journal (Theory of Computing): (i.) maintain zero cash flow, (ii.) adopt the internet as the primary means of dissemination and (even more radically) (iii.) leave copyright with the authors. See also the Electronic Journal of Combinatorics.
It is interesting to note that in order for this to be possible, not only does dissemination have to be free, but authors must typeset their own articles. If you are a member of the math/cs/physics community, you have been doing this yourself for perhaps 20 years.
Let us hope that a rapid cultural evolution divorces the publishing houses of these funds that the academic community can better spend on research and education!
A couple of comments about this.
:).
First of all, I have to say that if history is any guide, this proof is wrong. Incorrect claimed proofs of the twin prime conjecture outnumber correct proofs by a rather large margin
To calibrate this claim a bit it is worth noting that, to date, no one has even been able to prove that there are infinitely many primes pairs within distance o(log n); it has actually been proven for a constant c less than 1, there are finitely many pairs withing distance c log n, which I think is quite dramatic---I don't remember if this requires the Riemann Hypothesis or not.
(By the way, one say that there are infinitely many prime pairs withing distance f(n) of each other if there are infinitely many pairs (a_i, b_i) (both prime) so that |a_i - b_i | max(f(a_i),f(b_i)).)
Browsing through the above comments on this topic leaves me thoroughly discouraged with the Slashdot readership. SCO is *obviously* to blame; they probably thought the locks constituted IP infringement and arranged to pull the plug.
Presumably the marketing/bottom-line folks at RH have carefully considered the decision to cease boxed set sales. There *is* something romantic about the boxed sets, but I can understand their decision.
What does seem surprising (suspicious?) is the decision to leave package maintenance to the "developers themselves," without a period when the packages are under RH's control. This sounds a bit untenable, considering that RH's major value added feature is a distribution in which the (many) various components have undergone at least some kind of testing. I can't help but wonder how they will manage this.
Of course, opening up the development lists to the public seems like a positive move.
I guess we'll just have to wait until tomorrow to see what the details are...
The most puzzling aspect of this story is that the job of mapping the US internet is sufficient to earn a Ph.D. Of course, it is possible that there are aspects of the author's thesis that go beyond what is advertised above.
I admit that this author is not alone--in the CS department where I work, "experimental" Ph.D. theses featuring poorly designed experiments or no scientific work at all (which appears to be the case above) are a constant problem.
Perhaps this is an accident of the youth of the field.
What is being advertised here is not unbreakable in the sense used by most mathematician or serious cryptographers. (When a cryptographer says unbreakable, s/he means that the system is secure even against an adversary with unlimited computing power.)
Ideal use of a one time pad does have this property. There was a nice breakthrough in the EuroCrypt conference last year, where it was shown that one can obtain similar behavior even with keys that are shorter than the message to be encrypted, as long as the messages that you wish to encrypt are fairly random.
In any case, if you'd like to really understand what is going on here, for goodness' sake don't bother with Schneier's book; have a look at Goldreich's, "Foundations of Cryptography".