OK this is not a perfect example because of the Beatles copyrighted legacy, but sooner or later there's going to be one, two, three, fifty songs that get really popular but are only available online by independent vendors/artists.
How are they going to react then? They'd have lost before even starting. And eventually it will happen. It could have happened long ago, but what it needs is enough people sufficiently pissed off to go and create or find their own music. A portal that gets rid of distribution and merely links to media in an ordered way would be nice. (There probably are some?)
Remember mp3.com? I'll tell you what really killed it: it had WAAY to much quality music. It showed the emperor really is butt naked and not that pretty after all.
I had such a hangover from mp3.com that I haven't looked into online (independant) music for a while. Any suggestions?
"Until Feb 25th, I am going to be extremely busy with my wedding and honeymoon. I will be slow replying to non-wedding related emails during this time so please accept my apologies in advance. I expect to have a backlog of mail when I return so give me a few days to respond to these (probably by early March)."
Now, I hope his honeymoon was short, not his marriage. Perhaps they have an OpenMarriage though:)
Governments are not stupid. They may in fact be a lot more knowledgeable than, say, some fruit who thinks he's a journalist writing populist drivel at a MS shil site.
I'm sure they have some technically competent advisors. And then they have beancounters who make the very end decision cos in the end its all about the buck, not the bug.
Both authors are merely preaching to their choirs, it won't impact any real govt decision.
Yes, and I argued that one of the reasons (apart from this experiment being way too specific and time consuming) that this can occur is "ham rot", and the same can be got with "spam rot".
I'm not saying such an approach won't succeed, I'm saying that it's in fact inherent to how the filtering works and that you (with present techniques) are best off having a fresh ("naive") database and you shouldn't overtrain on the same stuff over and over because that will make your "unsure" margin all the larger.
I did read TFA. It wasn't all that specific. I didn't hear the audio.
Well, actually this appears to be one of the remarkable things that Bayesian systems showed: people suck at it. The problem I think is, that we tend to induldge in patterns we can interpret and the better we can the more valuable they must be (e.g. say, the reply-to address), and the first results with Bayesian experiments showed that patterns which seem meaningless to us -- hence: "naive filtering" -- often were just as or even more important statistically. I found this quite a surprise.
I do agree of course that more research/experiment on either side will make everything become more complex (perhaps up to the point where the whole thing becomes next to useless eventually).
"Bayesian depends on constistency in patterns" and "... if it looks for white fonts..."
It doesn't. It doesn't interpret anything. It doesn't look for words and certainly not for font colors. It merely remembers a bunch of tokens and whether or not you corrected (trained) the filter on unknowns and false positives.
If you roll your own spamfilter based on (human) interpretation or expectation or perceived correlation, I wish you best of luck but it would at best produce results similar to what you'd get with a personal Bayesian filter. Only instead of simple training you'd hardcode lots of rules that you *expect* to cover patterns you *think* you notice.
So, yes, I can understand why you're interested in rolling your own and keeping it secret, but no, I don't agree that it will be helpful.
It seems to me that "the bayesian filter authors stop processing those blocks" isn't really possible because of the binary nature of the filter and the total absence of any interpretation of the tokens being processed, for example in which combinations do they occur.
AFAIK this is not being done yet but it could provide a "context sensitive" weight factor for your tokens. It would likely make the math a lot more complex though and may not be practical because of CPU use and larger databases.
BTW, spammy hams should pose the same potential breakdown of the currently used - rather simple - Bayesian algorithms.
It's similar in the sense that what you want to achieve is that the spam/ham decision is as fuzzy as possible. Put another way: get those spam/ham graphs to be less steep or, taking a Bell curve as a model, get a very broad bell instead of a sharp peak. At the very least you'll increase the amount of "unsures" and false positives and thereby increase the need for training.
I'm not a statistics person, do I understand this correctly?
What matters is that while one person's spam might be very similar to another person's spam, their ham isn't. At best, it would require a semi-personal approach to sneak in spam. That's why you need to continually train your filter in the first place. Rinse and repeat, that's what it's all about.
What's being described is not really a flaw, but rather a saturation point at which it's time to retrain your filter and perhaps even start over with a new database. The old one gets too much 'noise' after some time.
They do point out one thing, be it from the spammers POV: Bayesian filtering is a continuous process and not and end to all solution. It requires fresh input and gets less effective if you keep old crud around for too long and if you train it too much on virtually the same spam/ham.
It's still a much better solution than blacklists.
I pronounce it 'skoh', their employees are said to do so too (or did). I always assumed everyone says 'Skoh'. It suits them.
Until a few days ago the MyDoom herring got in the news here (in Holland) and I learnt that their PR people and the parro^H^H^H^Hmedia seem to think it's 'Ess See Oh'.
But don't dispair, history will merely render them as the funny story that brought us the expression "pulling a sco".
"Tell me again about Skoh gramps";-)
"The Bauhaus school of software design"
on
FreeBSD 5.2 Review
·
· Score: 1
Hehe. Beautiful!
[Desktop use] Well let me put it this way
on
FreeBSD 5.2 Review
·
· Score: 4, Interesting
Works For Me (tm)
Desktop performance is a lot better on 5.x. Things like flash and other (binary linux) plugins actually work. Do use SCHED_ULE. It helps. Mplayer does it all, media plugins largely work. many of these issues are really external from FreeBSD but its nice to see things come together. Yes, you may have to fiddle a bit.
But it can be used on the desktop and it can work very well there. Like I said, things are starting to come together. Sometimes it looks like merely cosmetics from the Linux side I guess but as desktop apps get more mature so does their portability. Or at least easier to fix in ports. More hands and brains also help. There's clearly an influx into the BSD users realm.
So yes, there is a viable *BSD desktop other than Apple's (perhaps even 3 or 4). A true *NIX head or someone willing to read some docs can have a pretty complete desktop on top of a *BSD. I get GL animated snapshots from camera/tv card snapshots in my xscreensaver. Does windows have that?;-)
The writer argues that the BSDL "doesn't protect the rights of end-users the way the GPL does because it does not require the publisher to make the source code available". I don't get this. Unless one would s/end-users/the-code. And I have never understood what giving freedoms to a work means.
In other words, the writer is suggesting something to distract from the real point which is at the heart of the controversy BSDL vs GPL: whose freedom and freedom in the sense of "freedom to (do...)" or "freedom from (other entity doing...)".
To argue that the fact that BSDL code can be incorperated into a proprietary product is somehow an attack on the rights of the end user of *that* BSDL code certainly doesn't stand if one thinks about it for five seconds.
So it's the freedom of "the code" itself then? Please. Don't even *try* to make that argument.
Or the freedom to give something away with strings attached. There's nothing wrong with that, but then one shouldnt represent it as if it has any other meaning. Giving something away with no strings attached would somehow inherently be less of a contribution to society?
I have nothing against GPL personally but I do take offense at the ways its implications are time and time again used to discredit the BSDL with a completely reversed reasoning.
I think GPL is great for some things, linux kernel, gcc, and many more. BSD/MIT alike is more appropriate for other projects like apache, *BSD, and many more.
Look at GUI toolkits or the layers between toolkits and real focussed middleware. GPL does hamper the adoption of open source solutions (let alone development) there. Finance software for instance. So this is where (in terms of layers and libraries), BSD/MIT, or LGPL but thats a slippery one, makes sense. This is one (possibly not the most important, but it does count) reason for there being so much abandonware on sourceforge. People tend to slap a GPL license onto their work "because then it's free and not for MS".
Getting back to the GPL vs BSDL argument made, it's pretty clear that if you're feeling that someone else does something better you'd pour some moralism into your version of the difference in order to spin it your way. People should understand that if SCO is smart enough to understand how that works then RMS and his church certainly also are.
It's a delusion and yes it does prey on (often young) idealists providing them a world view just like any religion. There, I said it. Now, where's my protective suit.
Luckily many happy Linux users and developers realize this. But mod me down anyway.
- patience. "we" keep forgetting to be patient - focus. "we" got baffled with bullshit alright - history. "we" should have *started* with looking into the USL/BSDi case. That's where it'll end and it was very predictable. - agendas. "we" are stupid to follow the "ememy of enemy equals friend" idea. Distrust Novell. Distrust IBM. Trust me on this;-)
But most importantly, IMHO "we" forgot "our" role in all this. "We" played right into SCOs hand with the endless detail delving. No one but "us" cares. All they see is freaky commie geeks. Of course Darl knew and knows that. He made "us" jump on command. People have rightfully called him the comic relief but to most outsiders so are "we".
So what then? Here's something: sit and wait, perhaps document what happened when if you feel the need to. Write a book about it, someone will (please don't let that be JonKatz or ESR).
quote: "SCO, based in Lindon, Utah, is also sending letters to many of its 6,000 Unix licensees requiring them to certify in writing that they are complying with SCO licenses, a company executive said. SCO's Unix licensees are asked to certify that none of their employees or contractors have contributed any Unix code to Linux."
Truly mind boggling. I wonder if this is a good reason for a licensee to declare their contract with SCO void if SCO wants to add terms to it, and certainly if they are about use of ANOTHER OS.
I agree with the Win (and Mac?) ports of GTK sucking. Ppl who want crossplatform support usually go for wx instead I guess.
And (diffent subject, but related to what TP said last) I'm quite fond of GTK but I'm having a bad time with so much of GTK2 apps/functionality being tied into gnome. Guess what, I use xfce, not gnome but I get 3/4 of gnome nonetheless if I want a nice gtk2/AAfonts/themes desktop.
I've used GTK through py-GTK and wxPython (latter has better docs) and now that I'm using xfce as a DE for a while I have to say it has prettied up a lot. Having said that, qt/kde are leading here still and that's good for them.
You might have noticed I like python and unfortunately I've not managed to get py-qt/py-kde to build yet (FreeBSD) but I'm sure I'll have a look at that again soon once kde3.2 gets final and I want another DE to play with.
In case you're wondering, I've used mostly kde but also (on and off) blackbox, IceWM+ROX, xfce as DE. I never could stand gnome, but perhaps I should try that again too now everything (well almost, no?;-) is GTK2.
I don't like some of the FSF'ish zealots that seem to associate themselves with gnome either. And I can't but help to stress once more that the prime reason to start gnome was... kde. Slashback if you like but that was it.
And it's exactly the card that SCO is mostly playing, to get trenched in (former) bigshots to dilute the debate -- and that's only the debate on *this* side of the fence, on the other side I'm sure it goes quite differently and with lots of chuckling -- and step by step get more of their reasoning accepted or refuted quickly and then looked over, thus effectively accepted by the journaille.
The devil indeed is in the details. So they tell you to discuss details. Get it? They do. Not that that'll save them in the end. But surely they get that as well. This is their finest hour, or well at least Kevins;-)
Slashdot Mirror
OK this is not a perfect example because of the Beatles copyrighted legacy, but sooner or later there's going to be one, two, three, fifty songs that get really popular but are only available online by independent vendors/artists.
How are they going to react then? They'd have lost before even starting. And eventually it will happen. It could have happened long ago, but what it needs is enough people sufficiently pissed off to go and create or find their own music. A portal that gets rid of distribution and merely links to media in an ordered way would be nice. (There probably are some?)
Remember mp3.com? I'll tell you what really killed it: it had WAAY to much quality music. It showed the emperor really is butt naked and not that pretty after all.
I had such a hangover from mp3.com that I haven't looked into online (independant) music for a while. Any suggestions?
The guy just got married. From his website:
:)
"Until Feb 25th, I am going to be extremely busy with my wedding and honeymoon. I will be slow replying to non-wedding related emails during this time so please accept my apologies in advance. I expect to have a backlog of mail when I return so give me a few days to respond to these (probably by early March)."
Now, I hope his honeymoon was short, not his marriage. Perhaps they have an OpenMarriage though
Governments are not stupid. They may in fact be a lot more knowledgeable than, say, some fruit who thinks he's a journalist writing populist drivel at a MS shil site.
I'm sure they have some technically competent advisors. And then they have beancounters who make the very end decision cos in the end its all about the buck, not the bug.
Both authors are merely preaching to their choirs, it won't impact any real govt decision.
Yes, and I argued that one of the reasons (apart from this experiment being way too specific and time consuming) that this can occur is "ham rot", and the same can be got with "spam rot".
I'm not saying such an approach won't succeed, I'm saying that it's in fact inherent to how the filtering works and that you (with present techniques) are best off having a fresh ("naive") database and you shouldn't overtrain on the same stuff over and over because that will make your "unsure" margin all the larger.
I did read TFA. It wasn't all that specific. I didn't hear the audio.
Thanks for your answer. About pattern matching:
Well, actually this appears to be one of the remarkable things that Bayesian systems showed: people suck at it. The problem I think is, that we tend to induldge in patterns we can interpret and the better we can the more valuable they must be (e.g. say, the reply-to address), and the first results with Bayesian experiments showed that patterns which seem meaningless to us -- hence: "naive filtering" -- often were just as or even more important statistically. I found this quite a surprise.
I do agree of course that more research/experiment on either side will make everything become more complex (perhaps up to the point where the whole thing becomes next to useless eventually).
"Bayesian depends on constistency in patterns" and "... if it looks for white fonts ..."
It doesn't. It doesn't interpret anything. It doesn't look for words and certainly not for font colors. It merely remembers a bunch of tokens and whether or not you corrected (trained) the filter on unknowns and false positives.
If you roll your own spamfilter based on (human) interpretation or expectation or perceived correlation, I wish you best of luck but it would at best produce results similar to what you'd get with a personal Bayesian filter. Only instead of simple training you'd hardcode lots of rules that you *expect* to cover patterns you *think* you notice.
So, yes, I can understand why you're interested in rolling your own and keeping it secret, but no, I don't agree that it will be helpful.
Thanks for your answer.
It seems to me that "the bayesian filter authors stop processing those blocks" isn't really possible because of the binary nature of the filter and the total absence of any interpretation of the tokens being processed, for example in which combinations do they occur.
AFAIK this is not being done yet but it could provide a "context sensitive" weight factor for your tokens. It would likely make the math a lot more complex though and may not be practical because of CPU use and larger databases.
BTW, spammy hams should pose the same potential breakdown of the currently used - rather simple - Bayesian algorithms.
Yes this method is better.
It's similar in the sense that what you want to achieve is that the spam/ham decision is as fuzzy as possible. Put another way: get those spam/ham graphs to be less steep or, taking a Bell curve as a model, get a very broad bell instead of a sharp peak. At the very least you'll increase the amount of "unsures" and false positives and thereby increase the need for training.
I'm not a statistics person, do I understand this correctly?
Of course I can break my own Bayesian filtering.
What matters is that while one person's spam might be very similar to another person's spam, their ham isn't. At best, it would require a semi-personal approach to sneak in spam. That's why you need to continually train your filter in the first place. Rinse and repeat, that's what it's all about.
What's being described is not really a flaw, but rather a saturation point at which it's time to retrain your filter and perhaps even start over with a new database. The old one gets too much 'noise' after some time.
They do point out one thing, be it from the spammers POV: Bayesian filtering is a continuous process and not and end to all solution. It requires fresh input and gets less effective if you keep old crud around for too long and if you train it too much on virtually the same spam/ham.
It's still a much better solution than blacklists.
I pronounce it 'skoh', their employees are said to do so too (or did). I always assumed everyone says 'Skoh'. It suits them.
;-)
Until a few days ago the MyDoom herring got in the news here (in Holland) and I learnt that their PR people and the parro^H^H^H^Hmedia seem to think it's 'Ess See Oh'.
But don't dispair, history will merely render them as the funny story that brought us the expression "pulling a sco".
"Tell me again about Skoh gramps"
Hehe. Beautiful!
Works For Me (tm)
;-)
Desktop performance is a lot better on 5.x. Things like flash and other (binary linux) plugins actually work. Do use SCHED_ULE. It helps. Mplayer does it all, media plugins largely work. many of these issues are really external from FreeBSD but its nice to see things come together. Yes, you may have to fiddle a bit.
But it can be used on the desktop and it can work very well there. Like I said, things are starting to come together. Sometimes it looks like merely cosmetics from the Linux side I guess but as desktop apps get more mature so does their portability. Or at least easier to fix in ports. More hands and brains also help. There's clearly an influx into the BSD users realm.
So yes, there is a viable *BSD desktop other than Apple's (perhaps even 3 or 4). A true *NIX head or someone willing to read some docs can have a pretty complete desktop on top of a *BSD. I get GL animated snapshots from camera/tv card snapshots in my xscreensaver. Does windows have that?
Shun it, ignore it, cheer at it, but don't ever apologise for it unless perhaps if you created it. Submissive, misguided fools.
Sorry, I forgot to mention this. Credit where credit is due.
The writer argues that the BSDL "doesn't protect the rights of end-users the way the GPL does because it does not require the publisher to make the source code available". I don't get this. Unless one would s/end-users/the-code. And I have never understood what giving freedoms to a work means.
...)" or "freedom from (other entity doing...)".
In other words, the writer is suggesting something to distract from the real point which is at the heart of the controversy BSDL vs GPL: whose freedom and freedom in the sense of "freedom to (do
To argue that the fact that BSDL code can be incorperated into a proprietary product is somehow an attack on the rights of the end user of *that* BSDL code certainly doesn't stand if one thinks about it for five seconds.
So it's the freedom of "the code" itself then? Please. Don't even *try* to make that argument.
Or the freedom to give something away with strings attached. There's nothing wrong with that, but then one shouldnt represent it as if it has any other meaning. Giving something away with no strings attached would somehow inherently be less of a contribution to society?
I have nothing against GPL personally but I do take offense at the ways its implications are time and time again used to discredit the BSDL with a completely reversed reasoning.
I think GPL is great for some things, linux kernel, gcc, and many more. BSD/MIT alike is more appropriate for other projects like apache, *BSD, and many more.
Look at GUI toolkits or the layers between toolkits and real focussed middleware. GPL does hamper the adoption of open source solutions (let alone development) there. Finance software for instance. So this is where (in terms of layers and libraries), BSD/MIT, or LGPL but thats a slippery one, makes sense. This is one (possibly not the most important, but it does count) reason for there being so much abandonware on sourceforge. People tend to slap a GPL license onto their work "because then it's free and not for MS".
Getting back to the GPL vs BSDL argument made, it's pretty clear that if you're feeling that someone else does something better you'd pour some moralism into your version of the difference in order to spin it your way. People should understand that if SCO is smart enough to understand how that works then RMS and his church certainly also are.
It's a delusion and yes it does prey on (often young) idealists providing them a world view just like any religion. There, I said it. Now, where's my protective suit.
Luckily many happy Linux users and developers realize this. But mod me down anyway.
Send it to your favorite congressman, err woman, err entity. Point at the words Linux and GPL.
It might be worth its bucks after all.
Well likely not this particular thing, but the whole fantasy/porn/dreams thing it's going to be the drug of the future.
Keeps people distracted like nothing else.
Ladies and gentlemen, we got him!
I for one think this is an interesting look. Thanks.
Here are some rather random thoughts:
;-)
- patience. "we" keep forgetting to be patient
- focus. "we" got baffled with bullshit alright
- history. "we" should have *started* with looking into the USL/BSDi case. That's where it'll end and it was very predictable.
- agendas. "we" are stupid to follow the "ememy of enemy equals friend" idea. Distrust Novell. Distrust IBM. Trust me on this
But most importantly, IMHO "we" forgot "our" role in all this. "We" played right into SCOs hand with the endless detail delving. No one but "us" cares. All they see is freaky commie geeks. Of course Darl knew and knows that. He made "us" jump on command. People have rightfully called him the comic relief but to most outsiders so are "we".
So what then? Here's something: sit and wait, perhaps document what happened when if you feel the need to. Write a book about it, someone will (please don't let that be JonKatz or ESR).
quote: "SCO, based in Lindon, Utah, is also sending letters to many of its 6,000 Unix licensees requiring them to certify in writing that they are complying with SCO licenses, a company executive said. SCO's Unix licensees are asked to certify that none of their employees or contractors have contributed any Unix code to Linux."
Truly mind boggling. I wonder if this is a good reason for a licensee to declare their contract with SCO void if SCO wants to add terms to it, and certainly if they are about use of ANOTHER OS.
Partly reply to TP, mostly general comments...
;-) is GTK2.
I agree with the Win (and Mac?) ports of GTK sucking. Ppl who want crossplatform support usually go for wx instead I guess.
And (diffent subject, but related to what TP said last) I'm quite fond of GTK but I'm having a bad time with so much of GTK2 apps/functionality being tied into gnome. Guess what, I use xfce, not gnome but I get 3/4 of gnome nonetheless if I want a nice gtk2/AAfonts/themes desktop.
I've used GTK through py-GTK and wxPython (latter has better docs) and now that I'm using xfce as a DE for a while I have to say it has prettied up a lot. Having said that, qt/kde are leading here still and that's good for them.
You might have noticed I like python and unfortunately I've not managed to get py-qt/py-kde to build yet (FreeBSD) but I'm sure I'll have a look at that again soon once kde3.2 gets final and I want another DE to play with.
In case you're wondering, I've used mostly kde but also (on and off) blackbox, IceWM+ROX, xfce as DE. I never could stand gnome, but perhaps I should try that again too now everything (well almost, no?
I don't like some of the FSF'ish zealots that seem to associate themselves with gnome either. And I can't but help to stress once more that the prime reason to start gnome was... kde. Slashback if you like but that was it.
A statement "by an elder". That's what it is.
;-)
And it's exactly the card that SCO is mostly playing, to get trenched in (former) bigshots to dilute the debate -- and that's only the debate on *this* side of the fence, on the other side I'm sure it goes quite differently and with lots of chuckling -- and step by step get more of their reasoning accepted or refuted quickly and then looked over, thus effectively accepted by the journaille.
The devil indeed is in the details. So they tell you to discuss details. Get it? They do. Not that that'll save them in the end. But surely they get that as well. This is their finest hour, or well at least Kevins
With the slightly pissed off guy. "Use your UNIX skills on NT". It always pops up on every SCO story I've read. Always.
Surely something with chaos theory.
OK I should also say that I haven't exacly kept up with MS lately. Surely it's never been a buzzword though.
Greets