I like to think of analog as real numbers and digital as integers. Digital doesn't need to have only two values; it can have as many as you want, as long as the values are discrete.
So, you are basically correct wrt the way to increase data rates, but your use of analog of analog is misleading.
In principle, you can just increase the power to compensate, but the power required goes up exponentially once you hit your level spacing limit.
FWIW, you need to double the transmitter power (i.e., add 3dB more) to get an increase of 1 bit/sample. As an example, consider the common telephone line modem. The total power available to you, due to constraints by the telephone service, is very roughly around 30dB. QPSK modulation needs (very roughly) 9dB to transmit 2 bits/sample at a low enough error probability. Simple math says that we can then transmit around 9 bits/sample. Using a rough estimate of a 3000Hz bandwidth for the telephone channel, you get around 9 bits/sample * 1 sample.Hz * 3000 Hz = 27kbs, which is close to those 28.8kbs modems.
I should say that this is a very rough calculation. A lot of factors come into play when computing these rates. On one hand, the 3dB per bit/sample increase is only a estimate; it is actually a little less (i.e. the 21dB difference between what QPSK need and the 30dB available will get you more than just 7 bits). You can also use error correcting codes to increase a little bit more your rate. Lastly, the bandwidth may be a little higher than 3000Hz. On the other hand, it's really difficult to get 1 sample/Hz; it's more like 0.8 sample/Hz or even less.
IIRC, the 56kbs is achieved fiddling with the equipment interfacing with the local loop (i.e., the last mile) to get a higher bandwidth (but I may be wrong on this respect). This is done on the ISP side since it costs money to do that, and you'd have to do it for every phone line.
Note that you pay the price of a higher complexity to achieve this. To squeeze the 3000Hz bandwidth out of the telephone channel you need a good equalizer, and to achieve the 9 bits/sample you need complexity to implement the modulation (actually, most of the complexity is on the demodulator/decoder, i.e., on the receiver side). Old modems, e.g. the 9.6kbs, used a bandwidth of 2400Hz and a 16QAM modulation. The really old modems, i.e., the 300bps and 600bps used FSK, a modulation which is wasteful on bandwidth but yields simple receivers.
The result is that uninformed posts go straight to the top if they sound halfway reasonable. It's like peer review by a bunch of monkeys. Sure it's been reviewed, but that doesn't really give the content any more validity.
I think you are being unfair. It is true, as you pointed out, that a post needs only be a little reasonable to go to the top, and not all posts that go to the top are completely reasonable. However, this type of moderation already cuts dramaticaly the number of posts an informed reader would have to read to further solidify her own opinion. In other words, you are tacitly assuming that
3) Slashdot readers have no problems accepting without questioning comments moderators have marked as relevant.
which is definitely not true. I, for one, first read the article itself to form my opinion, and then try to read all the top posts to question it. That's, I think, the true spirit of/.
This method doesn't seem too secure to me. As long there is little ramdomness in the process, the end result is not secure, being susceptible to a brute force attack.
Let's see, you started with a common household phrase, which, depending on the length, doesn't contain too much randomness. You applied the eleet rule, which also is almost deterministic, and finalized with a deterministic permutation, the write-rows-read-columns. Knowing this whole procedure, it's just a matter of trying the most common phrases, applying this procedure before. Of course, if the attacker doesn't know the procedure, this is one source of randomness, but now that you've published it that randomness is gone.
As for my approach, I find it easier to let my brain learns the positions and sequence of the keys than to try to memorize the letter and hunt it down every time. I generate a random password using spwgen (available under debian) and then type it several times, trying to concentrate on the movement of the fingers, instead of the specific keys. So for example for the password 7$t-87c+ I try to concentrate on the fact that I use my right hand twice (with the left hand pressing the shift for the second key), and then my left followed again by the right 3 times etc. Of course this will depend on your own typing skills, but since for me it doesn't change that often, it makes for an easier to remember password.
I find I can remember quite a few of these passwords (even those I have not used for a while) and it makes learning new ones quite painless. Of course, I'd advise you to write them down for the first few days, keeping in a safe place (like pasted on your monitor;-). My favorite method of generating random passwords is, well, generating random passwords. I use a program called spwgen (available under Debian) to spit a few passwords, and then I type one at a time to find one that is too awkward to type. I then just type it several times to memorize the sequence of keys. I find I more readily remember the password if let the brain remember the
Slashdot Mirror
So, you are basically correct wrt the way to increase data rates, but your use of analog of analog is misleading.
I should say that this is a very rough calculation. A lot of factors come into play when computing these rates. On one hand, the 3dB per bit/sample increase is only a estimate; it is actually a little less (i.e. the 21dB difference between what QPSK need and the 30dB available will get you more than just 7 bits). You can also use error correcting codes to increase a little bit more your rate. Lastly, the bandwidth may be a little higher than 3000Hz. On the other hand, it's really difficult to get 1 sample/Hz; it's more like 0.8 sample/Hz or even less.
IIRC, the 56kbs is achieved fiddling with the equipment interfacing with the local loop (i.e., the last mile) to get a higher bandwidth (but I may be wrong on this respect). This is done on the ISP side since it costs money to do that, and you'd have to do it for every phone line.
Note that you pay the price of a higher complexity to achieve this. To squeeze the 3000Hz bandwidth out of the telephone channel you need a good equalizer, and to achieve the 9 bits/sample you need complexity to implement the modulation (actually, most of the complexity is on the demodulator/decoder, i.e., on the receiver side). Old modems, e.g. the 9.6kbs, used a bandwidth of 2400Hz and a 16QAM modulation. The really old modems, i.e., the 300bps and 600bps used FSK, a modulation which is wasteful on bandwidth but yields simple receivers.
I think you are being unfair. It is true, as you pointed out, that a post needs only be a little reasonable to go to the top, and not all posts that go to the top are completely reasonable. However, this type of moderation already cuts dramaticaly the number of posts an informed reader would have to read to further solidify her own opinion. In other words, you are tacitly assuming that
which is definitely not true. I, for one, first read the article itself to form my opinion, and then try to read all the top posts to question it. That's, I think, the true spirit ofLet's see, you started with a common household phrase, which, depending on the length, doesn't contain too much randomness. You applied the eleet rule, which also is almost deterministic, and finalized with a deterministic permutation, the write-rows-read-columns. Knowing this whole procedure, it's just a matter of trying the most common phrases, applying this procedure before. Of course, if the attacker doesn't know the procedure, this is one source of randomness, but now that you've published it that randomness is gone.
As for my approach, I find it easier to let my brain learns the positions and sequence of the keys than to try to memorize the letter and hunt it down every time. I generate a random password using spwgen (available under debian) and then type it several times, trying to concentrate on the movement of the fingers, instead of the specific keys. So for example for the password 7$t-87c+ I try to concentrate on the fact that I use my right hand twice (with the left hand pressing the shift for the second key), and then my left followed again by the right 3 times etc. Of course this will depend on your own typing skills, but since for me it doesn't change that often, it makes for an easier to remember password.
I find I can remember quite a few of these passwords (even those I have not used for a while) and it makes learning new ones quite painless. Of course, I'd advise you to write them down for the first few days, keeping in a safe place (like pasted on your monitor;-). My favorite method of generating random passwords is, well, generating random passwords. I use a program called spwgen (available under Debian) to spit a few passwords, and then I type one at a time to find one that is too awkward to type. I then just type it several times to memorize the sequence of keys. I find I more readily remember the password if let the brain remember the