"There is absolutely no chance that our hosts could have influenced the result and the suggestion is offensive. To Jiao Tong, to Bill Poucher, and to me."
You're about the 10th person to suggest that my concerns were related to cheating. I'd like to respectfully request that you appologize for putting those words in my mouth.
My summary of my discussion with Mr. Poucher has been posted as a sibling to your response.
"I respectfully disagree with your premise. Human female breasts are NEVER uninteresting!"
And a female pigeon would say, "I respectfully disagree with your premise. Pigeon male neck coloration is NEVER uninteresting."
Like I said, it's a display trait that has no particular value on its own as an evolutionary development compared to any other display trait such as long hair or fatty buttocks. The fact that they're capable of producing nurishment for under-developed (compared not non-mammals') young IS quite interesting in terms of the pressures that it relieves for the development time-frame.
"The only thing RMS could do would be to make the license more like BSD. It cannot put more restrictions on the user, since he/she could just the old license."
This is about the 10th time this week that someone has said "you're wrong because [x]" where [x] is exactly what I just said. I know this is Slashdot, but please let's occasionally read what we're replying to.
Quoth me earlier, "Oh, I'm not insinuating anything. I'm just saying that it's pretty seriously unlikely that the only team in the WORLD to solve all 8 problems would also be the host, and there is a very high probability that there's some other causal association."
I had a conversation with one of the contest organizers and he did quite a bit to resolve all of my concerns (WAY above-and-beyond, IMHO and he was very professional in our discussion). The concerns I had about jet-lag and other factors were clearly something that they had thought about, and while you can never 100% remove these from the results, they do move the contest every year and provide cultural events to incent longer stays prior to the contest.
Also, security as described was quite impressive.
As for the Chinese team answering more questions than anyone else, he pointed out that they used an interesting strategy here. When they got to the end, they and several other teams had solved 7, so they looked at what other teams had done and discovered that one of the questions remaining had been solved by several other teams. In a very smart move, they chose this problem, and completed it easily. Others chose harder problems and were unable to complete them.
My heartfelt congratulations to the winners, and my compliments to those who organize this event. It's all quite impressive. To those who insisted on refering to my concerns as "conspiracy theories"... well, it's Slashdot, what can you do.
"When they say 'retroactive' they mean the users have the option of choosing which version of the GPL they agree to."
I know that's how it works, but that's not what the Sleepycat author says. He's talking about something that I can't imagine makes any sense, or could be enforced. That said, I've seen some people come up with loopholes in MUCH shorter legal documents.
Still, I agree fundamentally. I think the most important point here is that there's not GPLv3, and the guy talking about this IS NOT in charge of making one. Moglen will speak when he's ready, and THEN we can decide if the sky is falling or not.
No one can force a company to abide by the rules of a new version of the GPL for software the currently have under an old version. They can CHOOSE to apply the new version IF the author used the standard boilerplate license notice, but they can also CHOOSE not to
The article specifically states that there is no GPLv3 and they're not officially comenting on what they are considering for it when it does come out
The guy commenting is saying what he would like to see so that he can drop his unique license. That's fine, but it's not official word
If this were to happen in the doomsday sense, everyone woudl immediately fork old copies of the programs that they have the option to apply the GPLv2 to, and continue to maintain and license them as such.
Right, so for example, a friend of mine works for a financial firm which he reports makes use of (and even contributes to) OSS projects. That firm provides a service based on these products (both electronic and off-line transactions that they perform as part of their core business). If they attempt to make this retroactive, I assure you that the world will come crashing down on the FSF. Thousands of firms around the country will sue them OVER NIGHT, and honestly, I'd be more likely to donate to the defense of those firms than the FSF (regardless of the fact that I'm a huge fan of the GPLv2).
This is deeply irresponsible. Any project that ships software under the GPL is going to be spinning their wheels for months over this, and the Microsofts of the world just got a huge weapon to use against OSS usage. After all, now they can say that using GPL software not only costs you in terms of the usual TCO metrics, but there's a potentially hidden and as-yet-unknown cost that can be applied to retroactively!
"And talk about missing options sheesh! Best evolutionary adaptation? I vote breasts!"
Are you referring to the original development of the lactating teat or the exaggerated secondary sexual characteristic in adult human females?
The latter is just a display trait, and other than the interesting matter of being tied to human females being effectively in a permanent state of heat (not sure if this is unique among mammals, but I know it's at least quite rare), it's fairly uninteresting.
The lactating teat on the other hand is quite a remarkable development, and while I'm not sure I'd put it up there with language, you could make the argument that things like language are possible BECAUSE of the developments (like this one) which allow the young to experience a prolonged development stage outside of the womb. This prolonged development in turn makes the development of a more complex brain far more practical.
So, I half agree with you, they're pretty darned important, though I consider the reduced number of young and proportionally reduced number of teats on primates to be a bit of a step backward...
This is nothing new. Interpol has been using a system like this for about a decade to track organized crime. Hell, the software was demonstrated on television that long ago, so imagine how long such systems have REALLY been in place.
It doesn't take a tinfoil hat to realize that law enforcement's response to high technology MUST be to absorb it at quickly as possible and use it to further their goals... only their goals remain in question.
We tried using rechargables where I work for pagers. Problem was that the rechargables don't die the same way as regular batteries. Time and time again the on-call person would find that their pager had just "shut off", where with traditional batteries, you got several hours or even days of warning that the battery was low.
"Perhaps you should read into your OWN posting and think again."
Yeah, I did. No mention of conspiracy theories. No mention of any of the nutty things you're accusing me of. I'm just calmly and reasonably pointing out that this is a strong corolation, that in the interests of good sportsmanship should ALWAYS be investigated (yes, if it were held at the Bermuda Community College, and they answered all 8 questions when no one else did, I'd suggest looking into any possible bias there too).
"The same example could be applied to the Olympics."
Oh most certainly, and it HAS. Keeping in mind that I suggested no cheating or conspiracy, we can liken this to Bob Beamon's controvercial long-jump in the 1968 Olympics. It has been suggested by many that the amazing distance he was able to cover was due to the reduced air-resistance at high altitude, and that his record should not be compared to lower-elevation games without re-calibration.
The same might apply here. There are thousands of reasons that one team might answer 8 questions where no one else was able to. Some are language-related (clearer translations in one language), cultural (the problems might have been ones which this school often deals with), geographical (I think this is a remotely administered test, but if not, jet-lag could be a factor), etc.
I'd really like to hear from someone who was involved, and get a sense of how they felt about the test and its administration.
Please read what I wrote rather than responding to what you WANT me to have written. I accused no one of anything. This is an anomoly, and a very pronounced one (they did not just WIN, they were the only team in the world to solve all of the problems).
I would tend to suspect a flaw in test administration due to language or cultural bias rather than outright cheating, myself (see, no conspiracy theory). Still, it's worth looking into.
More simply: You waste lots of energy when you turn on a florescent light, so turning them off, then having cleaning staff turn them on and then off, then having security turn them on and then off... that's going to burn FAR more energy than leaving them on.
Digital Equipment Corporation (now HP) once did a study along these lines. They found that not only did you use more power turning the lights off, but because of the cost of balast, and the fact that the single largest source of wear on balast was power-on, you also spent far more money replacing lights than you did powering them over night.
On all counts (energy usage, cost, etc), it's a win ot leave them on.
"As for the people who have been insinuating that the Shanghai Jiao Tong University rigged the results"
Oh, I'm not insinuating anything. I'm just saying that it's pretty seriously unlikely that the only team in the WORLD to solve all 8 problems would also be the host, and there is a very high probability that there's some other causal association. I don't want to insinuate that it was "cheating" per se (could be that a language, cultural or geophysical bias was introduced).
Still, it's a pretty strange thing. Hmmm... in fact, did anyone plot the results on a map? Are there any physical, political or cultural corollations? Does anyone who took part want to speak up? Were there any difficulties that you had other than those implicit in the test itself?
It's not shocking at all. First off, the only school IN THE WORLD to solve 8 problems in the allocated time was the host. That's too much of a coincidence to be ignored, so we'll discount their win as a potentially disputable data-point.
Second, the problems are hard. VERY HARD, so IMHO it comes down to more a matter of talent than of knowledge (we assume that students from each of the top-30 schools were given access to roughly the same information about the state of computer science techniques as far as they apply to the problem domains given, after all none of the problems requires knowledge of techniques newer than 20 years).
So it comes down to: how many of your students are interested in such a contest AND are capable of winning. MIT, CalTech, UI and the others who tied for 17th or 29th are strong schools with good students, but if you're going to take all of the strong schools with good students and find the smartest students among them... you'll probably find it's a roughtly even distribution, and this year's winners being in other countries is highly probable, though not certain.
Man, people just keep adding on a letter until their heads explode... what does the "L" stand for?
I think I prefer the term "open source", RMS's objections be damned, and no, I was not refering to open source software, I was refering to the GPL.
The GPL as a tool achieves a specific balance that companies tend to appreciate for their in-house projects. When they release the fruit of their work, they get a benefit: others work on it, improve on it, keep it current, etc. On the other hand, they're not handing a company a free ticket to productize their internal process. The worst-case scenario for many of these companies is having a consulting firm show up with some snazy new feature added, no source code for the change, and a $1,000,000 price tag for support of the entire app. That fear has killed many a project (even given the GPL, and the fact that it's not very likely).
Few things you may not know about NH: First off, its population is in an explosion. Especially around the southern border with MA, cities like Salem, Nashua, Derry and Manchester are thriving havens for high-tech businesses that don't want to cope with MA tax-rates.
Southern NH is also a favorite place for MA workers to buy a home because they are cheaper.
In 2000 they were the 41st most populace state, but 20th in terms of population density. However, they are also one of the most attractive vacation spots for both warm and cold weather in New England. How many Google users are going to want to see the ski slopes of various resorts to compare trail sizes? How many would like to see where National Park trails are in the White Mountains? How many MA shoppers cross the line into NH to shop tax-free (and need directions / arial views)?
NH is probably the second most traveled TO state in NE (CN being #1 due to its proximity to NY) unless you count the stop-over traffic in Logal Intl Airport (BOS).
"SO, I can buy a class-c and be able to spam you 256 times? (assuming, of course, only a single spam message per IP gets thru each time)."
Not at all.
You can SPAM 256 times, but it's almost certain that you won't spam ME 256 times. That's a key thing here that most folks don't get.
Let's take AOL as a random example. They took the hard-line approach of saying, "if we don't think you're serverish, we're not going to accept mail from you."
Had they, instead, said, "if we don't think you're serverish, we're going to greylist you," then they would only VERY rarely get spam from those systems. Greylisting is a wonderful tool if used carefully (and a horrible nightmare if used poorly, as I discovered when configuring it) because it doesn't really do much of anything about spam.
All greylisting does is reject mail for some configurable amount of time if its from and to addresses (in the envelope, not headers) do not match up with the sending IP in a historical database. In practice this means that your spammer with 256 addresses is going to try to send me spam, I'll reject him with a temporary error.
Assuming he's willing to actually try again later (most don't), he'll come back in a while, and I'll reject him again. Why? Because between when I rejected him the first time and now, he's touched dozens of honeypots and is now listed in abuse DNSBLs. This process works very well in practice, and I see a good fraction of those who get greylisted come back later and get rejected due to a DNSBL listing.
I'm sorry, but the idea that only blocking known offenders is unworkable has been proven wrong over and over.
I use a combination of greylisting, SPF and a small number of blacklists which have strict non-collateral damage policies.
Today, as an example, on a small personal system I've actively rejected 2576 connections, and allowed 228 messages. Of those 228, 75 were then identified as spam by SpamAssassin. A 97% success rate on a VERY low-bandwidth / CPU first-pass is more than acceptable for almost any application, given that you have a second pass (e.g. SA) which further improves your results to about the 99.9+% level.
The trap that people end up in is thinking that they need their first-pass to be as effective as a stand-along spam filter. Not true. You only need it to be effective enough to reduce the burden on your network and hardware by skimming off most of the incoming spam before it has a chance to consume those resources. If you're a VERY large ISP, then you might need to adopt additional measures (and while I despise the way AOL has done it, for example, I understand their reasons). If you're not one of the 10 largest ISPs in the world, then you are kidding yourself.
I have one user who asked me if mail was broken when I first deployed this. He was concerned because he'd come to think of the steady trickle of spam as a sort of heartbeat.
You should never trust any RBL, but if you must, you should pick one which defines a VERY narrow criteria with NO collateral damage.
Time and time again, I see people trying to enforce someone else's terms of service (usually poorly, and without room for any exception), getting blacklisted for non-spam activities (e.g. using a provider that hosts a spammer willingly), etc, etc.
These are attacks on the nature of the Internet as a network of peers.
Spamhaus does a very good job with XBL of listing just systems that are known zombies, relays, etc.
Combined with a decent offender-only list of bulk spam sources (I use dnsbl.antispam.or.id), you get excellent results, with few (none that I've been able to discover through analysis) false positives.
SpamAssassin, of course, makes this a moot point by combining and weighting several sources. I've never seen a false positive from SA as a result of bad blacklist handling (other tests, sure, but not it's DNSBLs). However, you may need some pre-filtering at SMTP time to reduce the load on your spam-filtering system, and that's where the above strategy comes back into play.
"Don't you just love it when the Linux Bigots think ecveryone should be in business to give away all they develop with their money."
Actually, the vast majority of those of us who are (and have been for nearing two decades) fans of the GPL are that way because we don't particularly care about software.
I'm a programmer, but I've only rarely worked for actual software companies. In most non-software companies, you hire programmers to make the things that off-the-shelf software doesn't provide possible.
For such efforts, the GPL is ideal, and I've seen companies benifit both from using established GPL projets as a starting point and from starting new GPL projects.
I told you so. I said it quietly, and never bothered a mailing list with it, that I recall, but anyone who brought up BK in front of me got a very serious "this is the end of the conversation, unless you want to hear my opinion on the damage BK is doing to open source."
There are several ways in which BK was a lose, but the one thing that always stuck out for me was their attitude toward their "customers". This idea that you could use their software, but you couldn't use it if you were potential competition. Even Microsoft has never stooped that low.
Slashdot Mirror
"So far this year we've had lots of advertisements for Thinkgeek"
Other than the running gag on April 1, what ads are you refering to? The banners, or actual stories?
"There is absolutely no chance that our hosts could have influenced the result and the suggestion is offensive. To Jiao Tong, to Bill Poucher, and to me."
You're about the 10th person to suggest that my concerns were related to cheating. I'd like to respectfully request that you appologize for putting those words in my mouth.
My summary of my discussion with Mr. Poucher has been posted as a sibling to your response.
"I respectfully disagree with your premise. Human female breasts are NEVER uninteresting!"
And a female pigeon would say, "I respectfully disagree with your premise. Pigeon male neck coloration is NEVER uninteresting."
Like I said, it's a display trait that has no particular value on its own as an evolutionary development compared to any other display trait such as long hair or fatty buttocks. The fact that they're capable of producing nurishment for under-developed (compared not non-mammals') young IS quite interesting in terms of the pressures that it relieves for the development time-frame.
"The only thing RMS could do would be to make the license more like BSD. It cannot put more restrictions on the user, since he/she could just the old license."
This is about the 10th time this week that someone has said "you're wrong because [x]" where [x] is exactly what I just said. I know this is Slashdot, but please let's occasionally read what we're replying to.
Quoth me earlier, "Oh, I'm not insinuating anything. I'm just saying that it's pretty seriously unlikely that the only team in the WORLD to solve all 8 problems would also be the host, and there is a very high probability that there's some other causal association."
I had a conversation with one of the contest organizers and he did quite a bit to resolve all of my concerns (WAY above-and-beyond, IMHO and he was very professional in our discussion). The concerns I had about jet-lag and other factors were clearly something that they had thought about, and while you can never 100% remove these from the results, they do move the contest every year and provide cultural events to incent longer stays prior to the contest.
Also, security as described was quite impressive.
As for the Chinese team answering more questions than anyone else, he pointed out that they used an interesting strategy here. When they got to the end, they and several other teams had solved 7, so they looked at what other teams had done and discovered that one of the questions remaining had been solved by several other teams. In a very smart move, they chose this problem, and completed it easily. Others chose harder problems and were unable to complete them.
My heartfelt congratulations to the winners, and my compliments to those who organize this event. It's all quite impressive. To those who insisted on refering to my concerns as "conspiracy theories"... well, it's Slashdot, what can you do.
"When they say 'retroactive' they mean the users have the option of choosing which version of the GPL they agree to."
I know that's how it works, but that's not what the Sleepycat author says. He's talking about something that I can't imagine makes any sense, or could be enforced. That said, I've seen some people come up with loopholes in MUCH shorter legal documents.
Still, I agree fundamentally. I think the most important point here is that there's not GPLv3, and the guy talking about this IS NOT in charge of making one. Moglen will speak when he's ready, and THEN we can decide if the sky is falling or not.
Right, so for example, a friend of mine works for a financial firm which he reports makes use of (and even contributes to) OSS projects. That firm provides a service based on these products (both electronic and off-line transactions that they perform as part of their core business). If they attempt to make this retroactive, I assure you that the world will come crashing down on the FSF. Thousands of firms around the country will sue them OVER NIGHT, and honestly, I'd be more likely to donate to the defense of those firms than the FSF (regardless of the fact that I'm a huge fan of the GPLv2).
This is deeply irresponsible. Any project that ships software under the GPL is going to be spinning their wheels for months over this, and the Microsofts of the world just got a huge weapon to use against OSS usage. After all, now they can say that using GPL software not only costs you in terms of the usual TCO metrics, but there's a potentially hidden and as-yet-unknown cost that can be applied to retroactively!
Grrr!
"And talk about missing options sheesh! Best evolutionary adaptation? I vote breasts!"
Are you referring to the original development of the lactating teat or the exaggerated secondary sexual characteristic in adult human females?
The latter is just a display trait, and other than the interesting matter of being tied to human females being effectively in a permanent state of heat (not sure if this is unique among mammals, but I know it's at least quite rare), it's fairly uninteresting.
The lactating teat on the other hand is quite a remarkable development, and while I'm not sure I'd put it up there with language, you could make the argument that things like language are possible BECAUSE of the developments (like this one) which allow the young to experience a prolonged development stage outside of the womb. This prolonged development in turn makes the development of a more complex brain far more practical.
So, I half agree with you, they're pretty darned important, though I consider the reduced number of young and proportionally reduced number of teats on primates to be a bit of a step backward...
This is nothing new. Interpol has been using a system like this for about a decade to track organized crime. Hell, the software was demonstrated on television that long ago, so imagine how long such systems have REALLY been in place.
It doesn't take a tinfoil hat to realize that law enforcement's response to high technology MUST be to absorb it at quickly as possible and use it to further their goals... only their goals remain in question.
We tried using rechargables where I work for pagers. Problem was that the rechargables don't die the same way as regular batteries. Time and time again the on-call person would find that their pager had just "shut off", where with traditional batteries, you got several hours or even days of warning that the battery was low.
"Perhaps you should read into your OWN posting and think again."
Yeah, I did. No mention of conspiracy theories. No mention of any of the nutty things you're accusing me of. I'm just calmly and reasonably pointing out that this is a strong corolation, that in the interests of good sportsmanship should ALWAYS be investigated (yes, if it were held at the Bermuda Community College, and they answered all 8 questions when no one else did, I'd suggest looking into any possible bias there too).
Please relax. Take a deep breath. It's ok.
"The same example could be applied to the Olympics."
Oh most certainly, and it HAS. Keeping in mind that I suggested no cheating or conspiracy, we can liken this to Bob Beamon's controvercial long-jump in the 1968 Olympics. It has been suggested by many that the amazing distance he was able to cover was due to the reduced air-resistance at high altitude, and that his record should not be compared to lower-elevation games without re-calibration.
The same might apply here. There are thousands of reasons that one team might answer 8 questions where no one else was able to. Some are language-related (clearer translations in one language), cultural (the problems might have been ones which this school often deals with), geographical (I think this is a remotely administered test, but if not, jet-lag could be a factor), etc.
I'd really like to hear from someone who was involved, and get a sense of how they felt about the test and its administration.
Please read what I wrote rather than responding to what you WANT me to have written. I accused no one of anything. This is an anomoly, and a very pronounced one (they did not just WIN, they were the only team in the world to solve all of the problems).
I would tend to suspect a flaw in test administration due to language or cultural bias rather than outright cheating, myself (see, no conspiracy theory). Still, it's worth looking into.
More simply: You waste lots of energy when you turn on a florescent light, so turning them off, then having cleaning staff turn them on and then off, then having security turn them on and then off... that's going to burn FAR more energy than leaving them on.
Digital Equipment Corporation (now HP) once did a study along these lines. They found that not only did you use more power turning the lights off, but because of the cost of balast, and the fact that the single largest source of wear on balast was power-on, you also spent far more money replacing lights than you did powering them over night.
On all counts (energy usage, cost, etc), it's a win ot leave them on.
"As for the people who have been insinuating that the Shanghai Jiao Tong University rigged the results"
Oh, I'm not insinuating anything. I'm just saying that it's pretty seriously unlikely that the only team in the WORLD to solve all 8 problems would also be the host, and there is a very high probability that there's some other causal association. I don't want to insinuate that it was "cheating" per se (could be that a language, cultural or geophysical bias was introduced).
Still, it's a pretty strange thing. Hmmm... in fact, did anyone plot the results on a map? Are there any physical, political or cultural corollations? Does anyone who took part want to speak up? Were there any difficulties that you had other than those implicit in the test itself?
It's not shocking at all. First off, the only school IN THE WORLD to solve 8 problems in the allocated time was the host. That's too much of a coincidence to be ignored, so we'll discount their win as a potentially disputable data-point.
Second, the problems are hard. VERY HARD, so IMHO it comes down to more a matter of talent than of knowledge (we assume that students from each of the top-30 schools were given access to roughly the same information about the state of computer science techniques as far as they apply to the problem domains given, after all none of the problems requires knowledge of techniques newer than 20 years).
So it comes down to: how many of your students are interested in such a contest AND are capable of winning. MIT, CalTech, UI and the others who tied for 17th or 29th are strong schools with good students, but if you're going to take all of the strong schools with good students and find the smartest students among them... you'll probably find it's a roughtly even distribution, and this year's winners being in other countries is highly probable, though not certain.
"FLOSS"
Man, people just keep adding on a letter until their heads explode... what does the "L" stand for?
I think I prefer the term "open source", RMS's objections be damned, and no, I was not refering to open source software, I was refering to the GPL.
The GPL as a tool achieves a specific balance that companies tend to appreciate for their in-house projects. When they release the fruit of their work, they get a benefit: others work on it, improve on it, keep it current, etc. On the other hand, they're not handing a company a free ticket to productize their internal process. The worst-case scenario for many of these companies is having a consulting firm show up with some snazy new feature added, no source code for the change, and a $1,000,000 price tag for support of the entire app. That fear has killed many a project (even given the GPL, and the fact that it's not very likely).
Few things you may not know about NH: First off, its population is in an explosion. Especially around the southern border with MA, cities like Salem, Nashua, Derry and Manchester are thriving havens for high-tech businesses that don't want to cope with MA tax-rates.
Southern NH is also a favorite place for MA workers to buy a home because they are cheaper.
In 2000 they were the 41st most populace state, but 20th in terms of population density. However, they are also one of the most attractive vacation spots for both warm and cold weather in New England. How many Google users are going to want to see the ski slopes of various resorts to compare trail sizes? How many would like to see where National Park trails are in the White Mountains? How many MA shoppers cross the line into NH to shop tax-free (and need directions / arial views)?
NH is probably the second most traveled TO state in NE (CN being #1 due to its proximity to NY) unless you count the stop-over traffic in Logal Intl Airport (BOS).
"SO, I can buy a class-c and be able to spam you 256 times? (assuming, of course, only a single spam message per IP gets thru each time)."
Not at all.
You can SPAM 256 times, but it's almost certain that you won't spam ME 256 times. That's a key thing here that most folks don't get.
Let's take AOL as a random example. They took the hard-line approach of saying, "if we don't think you're serverish, we're not going to accept mail from you."
Had they, instead, said, "if we don't think you're serverish, we're going to greylist you," then they would only VERY rarely get spam from those systems. Greylisting is a wonderful tool if used carefully (and a horrible nightmare if used poorly, as I discovered when configuring it) because it doesn't really do much of anything about spam.
All greylisting does is reject mail for some configurable amount of time if its from and to addresses (in the envelope, not headers) do not match up with the sending IP in a historical database. In practice this means that your spammer with 256 addresses is going to try to send me spam, I'll reject him with a temporary error.
Assuming he's willing to actually try again later (most don't), he'll come back in a while, and I'll reject him again. Why? Because between when I rejected him the first time and now, he's touched dozens of honeypots and is now listed in abuse DNSBLs. This process works very well in practice, and I see a good fraction of those who get greylisted come back later and get rejected due to a DNSBL listing.
This is a myth.
I'm sorry, but the idea that only blocking known offenders is unworkable has been proven wrong over and over.
I use a combination of greylisting, SPF and a small number of blacklists which have strict non-collateral damage policies.
Today, as an example, on a small personal system I've actively rejected 2576 connections, and allowed 228 messages. Of those 228, 75 were then identified as spam by SpamAssassin. A 97% success rate on a VERY low-bandwidth / CPU first-pass is more than acceptable for almost any application, given that you have a second pass (e.g. SA) which further improves your results to about the 99.9+% level.
The trap that people end up in is thinking that they need their first-pass to be as effective as a stand-along spam filter. Not true. You only need it to be effective enough to reduce the burden on your network and hardware by skimming off most of the incoming spam before it has a chance to consume those resources. If you're a VERY large ISP, then you might need to adopt additional measures (and while I despise the way AOL has done it, for example, I understand their reasons). If you're not one of the 10 largest ISPs in the world, then you are kidding yourself.
I have one user who asked me if mail was broken when I first deployed this. He was concerned because he'd come to think of the steady trickle of spam as a sort of heartbeat.
You should never trust any RBL, but if you must, you should pick one which defines a VERY narrow criteria with NO collateral damage.
Time and time again, I see people trying to enforce someone else's terms of service (usually poorly, and without room for any exception), getting blacklisted for non-spam activities (e.g. using a provider that hosts a spammer willingly), etc, etc.
These are attacks on the nature of the Internet as a network of peers.
Spamhaus does a very good job with XBL of listing just systems that are known zombies, relays, etc.
Combined with a decent offender-only list of bulk spam sources (I use dnsbl.antispam.or.id), you get excellent results, with few (none that I've been able to discover through analysis) false positives.
SpamAssassin, of course, makes this a moot point by combining and weighting several sources. I've never seen a false positive from SA as a result of bad blacklist handling (other tests, sure, but not it's DNSBLs). However, you may need some pre-filtering at SMTP time to reduce the load on your spam-filtering system, and that's where the above strategy comes back into play.
Hey, if Japan just gets me decent sat. imagery for New Hampshire that Google Maps can use, I'd be happy ;-)
"Don't you just love it when the Linux Bigots think ecveryone should be in business to give away all they develop with their money."
Actually, the vast majority of those of us who are (and have been for nearing two decades) fans of the GPL are that way because we don't particularly care about software.
I'm a programmer, but I've only rarely worked for actual software companies. In most non-software companies, you hire programmers to make the things that off-the-shelf software doesn't provide possible.
For such efforts, the GPL is ideal, and I've seen companies benifit both from using established GPL projets as a starting point and from starting new GPL projects.
I told you so. I said it quietly, and never bothered a mailing list with it, that I recall, but anyone who brought up BK in front of me got a very serious "this is the end of the conversation, unless you want to hear my opinion on the damage BK is doing to open source."
There are several ways in which BK was a lose, but the one thing that always stuck out for me was their attitude toward their "customers". This idea that you could use their software, but you couldn't use it if you were potential competition. Even Microsoft has never stooped that low.