But I still think algorithms should be public domain! If you own a company, and you have a particularly cool algorithm you want to hide, you should have to either obfuscate or encrypt the machine code.
Well it won't run very well if it is encrypted, and if you supply the key with the product, encryption won't help very much.
Your remark that algoritmes should be in the public domain means exactly that one could not make a living designing algorithmes.
That SCO also participated in Linux development is utterly irrelevant unless they themselves also put proprietary Unix code into Linux. Or put Linux code in Unix.
Actually, i would have see a wrong line producing the right MD5 sum once in every 340282366920938463463374607431768211456 lines, which is quite better then finding 2^16 of such for each line.
So the problem boils down to: Can you brute force a valid C line that matches a MD5SUM. If my assumption is correct that there will be many easy guesses of 5 lines (like a series of closing curly braces on on each line with tabs in between and so) and that a line of C carries an entropy of 6 bytes, my spare linux box should be able to spit out one line per day or so.. if only i had time to work out the details..;)
How long is your average c code line? A simple test on module.c and using gzip as an estimation for the number of relevant bits, i come to 6 relevant bytes per line. Not a number too large to crack.
The variance might make me stumble on a difficult line sometimes: A silly comment would be a nice salt value for that line. But then again i might find other easy starting points for some obvious block of 5 lines. A series of #includes would not be hard to guess. I won't volunteer to actually do this, but i don't think it is cryptographically safe either.
It seems to me the proposed use of MD5 sums is quite dangerous here, meaning that the md5 sum list might give away the original source.
It won't be hard to find the source matching the first 5 lines. probably somethine with/*, the words copyright and SCO and such.
The next line can then be guessed, and this will be relatively easy because it has to be valid C. I don't think there is much 'randomness'in C, so it should be fairly easy. Fairly easy meaning with the use of brute force, which seems quite an appropriate term in this matter.
first they gave us culture, then they gave us linux, now they give us some common sense
Having seen what you've been doing with the culture, we now want it back.
Labor-saving devices in particular, the sort the article author derides most, are what give us the free time to read his articles about the uselessness of those devices. Which I find amusing.
And we have nice machines that do the hard work for us. So we save time and money, which we need for the gym to get some physical exercise...
Slashdot Mirror
On the A2 there is a 12 kilometer beowulf cluster.
Dann singe ich ein Lied fuer dich
Von 99 Luftballons
Bang! Von 98 Luftballons Bang! Von 97 Luftballons Bang! Bang!
But I still think algorithms should be public domain! If you own a company, and you have a particularly cool algorithm you want to hide, you should have to either obfuscate or encrypt the machine code.
Well it won't run very well if it is encrypted, and if you supply the key with the product, encryption won't help very much.
Your remark that algoritmes should be in the public domain means exactly that one could not make a living designing algorithmes.
That SCO also participated in Linux development is utterly irrelevant unless they themselves also put proprietary Unix code into Linux.
Or put Linux code in Unix.
Actually, i would have see a wrong line producing the right MD5 sum once in every 340282366920938463463374607431768211456 lines, which is quite better then finding 2^16 of such for each line. So the problem boils down to: Can you brute force a valid C line that matches a MD5SUM. If my assumption is correct that there will be many easy guesses of 5 lines (like a series of closing curly braces on on each line with tabs in between and so) and that a line of C carries an entropy of 6 bytes, my spare linux box should be able to spit out one line per day or so.. if only i had time to work out the details..;)
How long is your average c code line? A simple test on module.c and using gzip as an estimation for the number of relevant bits, i come to 6 relevant bytes per line. Not a number too large to crack.
The variance might make me stumble on a difficult line sometimes: A silly comment would be a nice salt value for that line. But then again i might find other easy starting points for some obvious block of 5 lines. A series of #includes would not be hard to guess.
I won't volunteer to actually do this, but i don't think it is cryptographically safe either.
It seems to me the proposed use of MD5 sums is quite dangerous here, meaning that the md5 sum list might give away the original source. /*, the words copyright and SCO and such.
The next line can then be guessed, and this will be relatively easy because it has to be valid C. I don't think there is much 'randomness'in C, so it should be fairly easy.
It won't be hard to find the source matching the first 5 lines. probably somethine with
Fairly easy meaning with the use of brute force, which seems quite an appropriate term in this matter.
first they gave us culture, then they gave us linux, now they give us some common sense
Having seen what you've been doing with the culture, we now want it back.
Labor-saving devices in particular, the sort the article author derides most, are what give us the free time to read his articles about the uselessness of those devices. Which I find amusing.
And we have nice machines that do the hard work for us. So we save time and money, which we need for the gym to get some physical exercise...