Notwithstanding accepted wisdom espoused above, random words cannot defeat current statistical spam filters, and it is difficult to defeat such filters even if you have access to the algorithm and the recipient's mailbox.
John Graham-Cumming presented a talk Beating Bayesian Filters at the 2004 Spam Conference detailing these results. A video recording is available; alas, no paper.
In conducting a recent spam filter evaluation I observed (but did not report) that the statistical filter attacks were not particularly effective. The only attack that worked sometimes was to make the entire body of the message a current news item or joke, with only a URL linking to the spam payload.
There are conflicting reports about whether or not CD sales are down, and if so, whether file sharing is a causal factor. RIAA may well be fudging both so as to fake damages.
One should never accept a popular-press hearsay account of a research report. Unfortunately my library doesn't carry "Science Express" (an ancillary to the respected "Science") where the paper appears.
"Science Express" has its own paraphrasal of the paper
at its website but you have to pay for the full text. There is also a link to "supporting online material" that includes a free document describing some of the methds and results.
Subject to the caveat that I did not fork over the $$$ for the full article, I'd say the conclusions appear unremarkable. Humans raised in cultures that lack counting can't count beyond 3, and also can't express the concept. I see no experiment that indicates
causality between what I consider two aspects
of the same phenomenon.
Have you read or heard anything Moglen has said? Indeed, have you read the GPL? A conditional copyright license, broadly interpreted, is a contract. But it is enforced, not by contract law, which in the U.S. is a state law, but by copyright law, which is federal and subject to international treaties. In a legal sense, a contract is a binding agreement in which two parties exchange some valuable consideration. A license need not involve the exchange of anything; it may be unilaterally granted.
GPL does not contain a clause like "by distributing this code, you agree..."
Another license could well have such a clause, in which case you would be bound by contract law, because distributing would be accepting a consideration.
The authors of the GPL decided to avoid recourse to contract law, which suffers from jurisdictional issues and lack of international enforcement agreements.
Distributing the code consitutes accepting a consideration: the consideration is a license from the copyright holder for that distribution. So if the conditions attached to the license say "distributing this code consitutes acceptance of these conditions" and you distribute the code, you're committed. For example, if the license said, "by distributing this code, you agree to pay me $10 per copy downloaded," you could be sued for the $10 per copy.
As Moglen argues, the beauty of GPL (and the other licenses mentioned) is that it does not depend on this acceptance. If you accept it, you have to obey the conditions. If you don't accept it, distribution is illegal because you have no license at all. However you slice it (accepted or not accepted) it is illegal to distribute without obeying the conditions.
Don't forget that their cash value is encumbered by any outstanding judgements they may collect. It seems to me at $125K per hit for "willful infringement of copyright" (the same hammer RIAA use on teeny-boppers), judgement liabilities could easily exceed their asset value.
IBM has a separate counterclaim re GPL and violation of IBM's copyrights.
Their point in this request for summary judgement has nothing to do with GPL. All they are saying is that SCO can't expect IBM to keep it (JFS etc.) confidential while they (SCO) continue to publish it. As SCO has stated over and over, this particular claim has nothing to do with copyright, and therefore nothing to do with GPL. It has to do with confidentiality.
As you get older, your eyes' ability to focus (accommodation) diminishes. Your eyes will tend to their default (relaxed) focal length. I'm a bit near sighted (-2.25 diopters correction), so in my rocking chair my eyes will focus at about 18 inches. That's a convenient distance. It will enable me to read comfortably, to see my computer monitor, and to see distance well enough that I don't walk into walls.
Independent of the risks of failure, lasic surgery accelerates the deterioration of accommodation. I'd sooner not accelrate that process, and in the limit I'd prefer an 18 inch focal length. So I'll play the cards I was dealt.
We make wine, too. The mass-produced ones are plonk, but there are some very respectable whites and serious reds as well. Some are even available in U.S. stores. Look for "Henry of Pellham" (Ontario), "Chateau des Charmes" (Ontario), and "Tinhorn Creek" (B.C) to name a couple.
If you happen to be flying through Toronto, the duty free has a Chateau-des-Charmes Cab. Franc (1998) for $20 (CAD) that is worth trying.
The press have screwed this one up, and Slashdot is parroting their line.
The judged stayed the proceedings. The only "limit" is that SCO were invited to file for a preliminary injunction while the stay is pending. Many read the invitation as "put up or shut up." It is a major reach to put any SCO-positive spin on this ruling.
I'd like to remind readers that USL vs BSDI was resolved when the court denied a motion for preliminary injunction.
Since this article was written, high-power solid-state amps have become common. Phase Linear was the first brand to popularize high power, with 500 and 700 watt/channel stereo amplifiers.
Typical solid state amplifiers have increased in power an headroom to the point that you are unlikely to want to listen to them at clipping.
It is certainly true that some people like the coloration introduced by tube amps. Guitar players routinely treat tubes as musical instruments by overdriving them.
Another (non-disjoint) set of people enjoy the coloration and noise of vinyl recordings.
The bottom line is that you can make a digital recording of your favourite vinyl/tube/whatever golden-ears setup, and be unable to distinguish it from the original in controlled A/B comparisons.
If you want to color your music, use tubes. If you want high fidelity, don't.
I think the poster and the referenced profs. rather regret the need for students in a university.
Students are an essential component of the community of smart people to which I (the poster) refer. Part-time and evening programs present major challenges -- professors and students have to work much harder to form a community under these circumstances. From your account, I daresay both could have made more effort.
Perhaps the university to which you refer should not offer such a program. Perhaps you should have chosen a college. But students seek the presitige of a university degree, and universities can ill afford to pass up the tuition and other credits that accrue from offering such programs.
I hope that you find what you're looking for at your chosen university.
Universities are in the knowledge business. Delivering lectures is a small part of what a professor does, especially in a top university. For that matter, "the curriculum" is a small part of the value of a university education.
The university provides an environment in which smart people have the opportunity to interact with other smart people. A professor's main role is to create new knowledge and, by example, to encourage students to acquire and create knowledge, using as resources their professors, their peers, text books, labs, etc.
People who are keen on knowledge are naturally keen to share it, and that's where the value of being a member of a university, student or professor, comes in. Think of it as an intellectual support group.
Professors in general are glad of being relieved of routine lecturing. This frees them to do more research, and to interact with students in more fulfilling ways.
I went to Secunia's site and saw no reference to the study, even for a price.
I am very suspicious of outfits who describe their results with hyperbole like "dispels the myth" but fail to make their research report available for scrutiny.
It sounds like this study uses "retrospective" methodology, which is OK, but one must be very careful to assess the meaning and transferrability of such results. In general, retrospectives don't yield blockbuster evidence that "dispels myths."
Here, in contrast is proposed hypothesis for a forward-looking study: I predict that a larger fraction of Windows machines will be compromised in the next 12 months than machines running non-windows operating systems. That is, the probability that a given Windows machine will be the victim of a successful attack is higher than the corresponding probability for a non-windows machine.
We shall not respond to Mr. Zdziarski's attacks, except to identify
the most outstanding factual errors and to note that ad hominem
arguments are irrelevant in assessing the validity of our work.
First, I'd like to thank you for your correspondence with me. The information you provided me with was very helpful.
I believe that you said the number of ham errors and spam errors should be roughly equal. They are. It is the proportions that are unequal, because there is much more spam than ham in the test suite.
An interesting issue arose in our correspondence as to how to introduce bias into CRM-114 so it would place more emphasis on ham accuracy. My understanding is that this is not an easy task, and obvious approaches like double-training don't work.
Slashdot Mirror
John Graham-Cumming presented a talk Beating Bayesian Filters at the 2004 Spam Conference detailing these results. A video recording is available; alas, no paper.
In conducting a recent spam filter evaluation I observed (but did not report) that the statistical filter attacks were not particularly effective. The only attack that worked sometimes was to make the entire body of the message a current news item or joke, with only a URL linking to the spam payload.
Here's a report that says "Nielsen Rating System At Odds With RIAA's Claim Of Lost Sales".
Here's a report that states "downloads have an effect on sales which is statistically indistinguishable from zero".
I don't know how much they paid, but here's an example of a formerly-good service being bought by Microsoft and replaced by crap: Mapblast.com.
Now that they're public, Microsoft can buy them and replace their service with whatever they choose.
"Science Express" has its own paraphrasal of the paper at its website but you have to pay for the full text. There is also a link to "supporting online material" that includes a free document describing some of the methds and results.
Subject to the caveat that I did not fork over the $$$ for the full article, I'd say the conclusions appear unremarkable. Humans raised in cultures that lack counting can't count beyond 3, and also can't express the concept. I see no experiment that indicates causality between what I consider two aspects of the same phenomenon.
It sounds like the judge read and took to heart Larry Lessig's book, Free Culture. A free download and a must-read!
Have you read or heard anything Moglen has said? Indeed, have you read the GPL? A conditional copyright license, broadly interpreted, is a contract. But it is enforced, not by contract law, which in the U.S. is a state law, but by copyright law, which is federal and subject to international treaties. In a legal sense, a contract is a binding agreement in which two parties exchange some valuable consideration. A license need not involve the exchange of anything; it may be unilaterally granted.
GPL does not contain a clause like "by distributing this code, you agree ..."
Another license could well have such a clause, in which case you would be bound by contract law, because distributing would be accepting a consideration.
The authors of the GPL decided to avoid recourse to contract law, which suffers from jurisdictional issues and lack of international enforcement agreements.
I agree 100%. I'm tired of hearing him dismissed as a Zealot by those who avail themselves of the fruit from the tree he planted.
Distributing the code consitutes accepting a consideration: the consideration is a license from the copyright holder for that distribution. So if the conditions attached to the license say "distributing this code consitutes acceptance of these conditions" and you distribute the code, you're committed. For example, if the license said, "by distributing this code, you agree to pay me $10 per copy downloaded," you could be sued for the $10 per copy.
As Moglen argues, the beauty of GPL (and the other licenses mentioned) is that it does not depend on this acceptance. If you accept it, you have to obey the conditions. If you don't accept it, distribution is illegal because you have no license at all. However you slice it (accepted or not accepted) it is illegal to distribute without obeying the conditions.
I think you're confusing "in court" with "before a jury". IBM is most certainly causing the GPL to be considered by the court.
Don't forget that their cash value is encumbered by any outstanding judgements they may collect. It seems to me at $125K per hit for "willful infringement of copyright" (the same hammer RIAA use on teeny-boppers), judgement liabilities could easily exceed their asset value.
IBM has a separate counterclaim re GPL and violation of IBM's copyrights.
Their point in this request for summary judgement has nothing to do with GPL. All they are saying is that SCO can't expect IBM to keep it (JFS etc.) confidential while they (SCO) continue to publish it. As SCO has stated over and over, this particular claim has nothing to do with copyright, and therefore nothing to do with GPL. It has to do with confidentiality.
As you get older, your eyes' ability to focus (accommodation) diminishes. Your eyes will tend to their default (relaxed) focal length. I'm a bit near sighted (-2.25 diopters correction), so in my rocking chair my eyes will focus at about 18 inches. That's a convenient distance. It will enable me to read comfortably, to see my computer monitor, and to see distance well enough that I don't walk into walls.
Independent of the risks of failure, lasic surgery accelerates the deterioration of accommodation. I'd sooner not accelrate that process, and in the limit I'd prefer an 18 inch focal length. So I'll play the cards I was dealt.
The real Sal jr. writes only in italics. Also, your spelling and grammar are too good.
Michael Moore's Canadian Bacon provides a hilarious contrast between U.S. and Canadian cliches. Starring Alan Alda and John Candy.
We make wine, too. The mass-produced ones are plonk, but there are some very respectable whites and serious reds as well. Some are even available in U.S. stores. Look for "Henry of Pellham" (Ontario), "Chateau des Charmes" (Ontario), and "Tinhorn Creek" (B.C) to name a couple.
If you happen to be flying through Toronto, the duty free has a Chateau-des-Charmes Cab. Franc (1998) for $20 (CAD) that is worth trying.
The press have screwed this one up, and Slashdot is parroting their line.
The judged stayed the proceedings. The only "limit" is that SCO were invited to file for a preliminary injunction while the stay is pending. Many read the invitation as "put up or shut up." It is a major reach to put any SCO-positive spin on this ruling.
I'd like to remind readers that USL vs BSDI was resolved when the court denied a motion for preliminary injunction.
Typical solid state amplifiers have increased in power an headroom to the point that you are unlikely to want to listen to them at clipping.
It is certainly true that some people like the coloration introduced by tube amps. Guitar players routinely treat tubes as musical instruments by overdriving them.
Another (non-disjoint) set of people enjoy the coloration and noise of vinyl recordings.
The bottom line is that you can make a digital recording of your favourite vinyl/tube/whatever golden-ears setup, and be unable to distinguish it from the original in controlled A/B comparisons.
If you want to color your music, use tubes. If you want high fidelity, don't.
Students are an essential component of the community of smart people to which I (the poster) refer. Part-time and evening programs present major challenges -- professors and students have to work much harder to form a community under these circumstances. From your account, I daresay both could have made more effort.
Perhaps the university to which you refer should not offer such a program. Perhaps you should have chosen a college. But students seek the presitige of a university degree, and universities can ill afford to pass up the tuition and other credits that accrue from offering such programs.
I hope that you find what you're looking for at your chosen university.
The university provides an environment in which smart people have the opportunity to interact with other smart people. A professor's main role is to create new knowledge and, by example, to encourage students to acquire and create knowledge, using as resources their professors, their peers, text books, labs, etc.
People who are keen on knowledge are naturally keen to share it, and that's where the value of being a member of a university, student or professor, comes in. Think of it as an intellectual support group.
Professors in general are glad of being relieved of routine lecturing. This frees them to do more research, and to interact with students in more fulfilling ways.
I went to Secunia's site and saw no reference to the study, even for a price.
I am very suspicious of outfits who describe their results with hyperbole like "dispels the myth" but fail to make their research report available for scrutiny.
It sounds like this study uses "retrospective" methodology, which is OK, but one must be very careful to assess the meaning and transferrability of such results. In general, retrospectives don't yield blockbuster evidence that "dispels myths."
Here, in contrast is proposed hypothesis for a forward-looking study: I predict that a larger fraction of Windows machines will be compromised in the next 12 months than machines running non-windows operating systems. That is, the probability that a given Windows machine will be the victim of a successful attack is higher than the corresponding probability for a non-windows machine.
Michael Moore doesn't own clear copyright to F911. So, while he may approve of sharing it, his comments do not consitute a license.
We encourage interested parties to read our paper and our points of fact re Zdziarski.
Thomas Lynam
Gordon Cormack
June 24, 2004
Bill,
First, I'd like to thank you for your correspondence with me. The information you provided me with was very helpful.
I believe that you said the number of ham errors and spam errors should be roughly equal. They are. It is the proportions that are unequal, because there is much more spam than ham in the test suite.
An interesting issue arose in our correspondence as to how to introduce bias into CRM-114 so it would place more emphasis on ham accuracy. My understanding is that this is not an easy task, and obvious approaches like double-training don't work.
GVC