The test was very insightfull and basically good. Sure enough it was not exhaustive nor objective, but it did cover the fundamental bits necessary to explain that most porn detection systems in general and the one described in specifics are mostly bogus.
A detection rate of 80% is just plain crap. OCR software which does not have a detection rate of 99%+ is not used because it is too much hassle to correct the >1% bugs. Surfing the web and arbitrarily getting block outs is just extremly annoying - I bet every ISP who dares to use this software will lose customers faster than AOL sends out new CDs...
More scientifically: Porn detection is just not possible at the moment. Even though the system presented has a reasonable success rate for a scientific conference (it would be rejected because the ideas are too old, not because they are stupid), it is not usable in real life. And no other system exists, neither theoretically nor practically.
Using Hash codes to identify images as suggested above is not really a solution since it is way too much effort to maintain such a system and it does not apply to the general case - and it might not even scale. I do not have a clue how many images are swimming around usenet, but I know that they do not fit onto a normal hard disk. It is just a question of time until neither the accumulated hash codes of these do fit on a hard disk either.
Yep, you could certainly do a lot with the money, but nevertheless you could not write a fairly complete, although difficult, ERP (Enterprise Resource Planning). SAP is in the market of ERP since more than ten years, and they did acquire some expertise in that field. It is definitely much cheaper to go with SAP R/3 than writing and maintaining and customizing and extending your own system...
First of, this is old news - the announcement was already out a couple of days ago, and SuSE 6.1 was already mentioned in a previous slashdot post...
Second of, SuSE is known for being late. Sadly, they partly follow the Microsoft strategy - announce early, deliver at some point. Going by experience, the "European" (I do not know why somebody calls a plain German version European...) version is going to be released at least two weeks late, and the international one will suffer in the same amount...
Not that I would claim that SuSE was / is first, but the 6.0 release of their distro came out in December and was "2.2 ready", at least that's what they claimed. The 6.1 release is supposed to come out middle of April with the 2.2.3 kernel.
I guess, SuSE as well as Caldera use the 2.2 as marketing hype... and who cares who is first - there is much more to a good distro than the newest kernel version.
Slashdot Mirror
The test was very insightfull and basically good. Sure enough it was not exhaustive nor objective, but it did cover the fundamental bits necessary to explain that most porn detection systems in general and the one described in specifics are mostly bogus.
A detection rate of 80% is just plain crap. OCR software which does not have a detection rate of 99%+ is not used because it is too much hassle to correct the >1% bugs. Surfing the web and arbitrarily getting block outs is just extremly annoying - I bet every ISP who dares to use this software will lose customers faster than AOL sends out new CDs...
More scientifically: Porn detection is just not possible at the moment. Even though the system presented has a reasonable success rate for a scientific conference (it would be rejected because the ideas are too old, not because they are stupid), it is not usable in real life. And no other system exists, neither theoretically nor practically.
Using Hash codes to identify images as suggested above is not really a solution since it is way too much effort to maintain such a system and it does not apply to the general case - and it might not even scale. I do not have a clue how many images are swimming around usenet, but I know that they do not fit onto a normal hard disk. It is just a question of time until neither the accumulated hash codes of these do fit on a hard disk either.
Yep, you could certainly do a lot with the money, but nevertheless you could not write a fairly complete, although difficult, ERP (Enterprise Resource Planning). SAP is in the market of ERP since more than ten years, and they did acquire some expertise in that field. It is definitely much cheaper to go with SAP R/3 than writing and maintaining and customizing and extending your own system...
First of, this is old news - the announcement was already out a couple of days ago, and SuSE 6.1 was already mentioned in a previous slashdot post...
Second of, SuSE is known for being late. Sadly, they partly follow the Microsoft strategy - announce early, deliver at some point. Going by experience, the "European" (I do not know why somebody calls a plain German version European...) version is going to be released at least two weeks late, and the international one will suffer in the same amount...
Not that I would claim that SuSE was / is first, but the 6.0 release of their distro came out in December and was "2.2 ready", at least that's what they claimed. The 6.1 release is supposed to come out middle of April with the 2.2.3 kernel.
I guess, SuSE as well as Caldera use the 2.2 as marketing hype... and who cares who is first - there is much more to a good distro than the newest kernel version.