Question - if a compression algorithm can ALWAYS reduce the length of the data, and produces data in the same format (binary in this case) as the input, then it can be applied recursively until nothing is left - the rulebook on the other side then applies the decompression algorithm until the original data is back... Of course, this is mathematically impossible, which is what I said before.

I've written a couple of codecs for distributed VST, I know what I'm talking about.
Not all compression uses dictionaries either - look at FLAC for example, which (if I remember correctly) is some variant of the FFT, as is MPEG, JPEG, DIVX

Psychogenesis,
As I said... I don't know *much about compression, but you are ignoring something that is a very intuitive and possible idea simply because you claim that it is mathematically "impossible" when I quite simply told you why it isn't impossible.  You are not reducing the "message" to zero you are reducing the message SENT to a small number.  For each level of decompression you have these chuncks of highly probability blocks already pre computed to do the reverse opperation.  These in and of themselves means you are not breaking any fundamantal rule you just "arelady have the answer". 

I misspoke, it isn't Turbo Pascal, it's Turbo Codes http://en.wikipedia.org/wiki/Turbo_codes.  Before they were published everyone had pretty much given up on the problem.  The authors used INTUITION to come up with the solution and THEN mathematically proved it to hold.  Your ignoring the intuition because your math is getting in your way.  Sure there may be degenerate cases where it expands something rather than compressing it, then... use another compression method or send the raw data.  No big deal.  Huffman code will also almost always compress a file, it *could be used recursively also.  There are degenerate cases for Huffamn also, does that mean ignore the entire possibility of Huffman codes?  Also Trubo codes [suprise suprise] use probabilities also.

From what I know about JPEG/MPEG they are lossy compression methods AND decompression is assisted by tables also that increase the compressibility of commonly occuring matrices.

Which I didn't deny.  We seem to be arguing about two different things.  I admitted that there are degenerate cases.  I believe I even said that in my very first post, btw arguing with you helped me solve another problem so thanks:D.  But you are ignoring the possibility of Einstein's method of *generally offering very very good compression for "most" data.  As data is inherentely skewed, so with sufficent study his method, especially if in the unlikely chance that no such method is "widely" published and knowing that many other "scholars" would view the claims similarly to what you and just about everyone else on this board did and has done [what was my point again?].  Since it probaby hasn't been published before then yes in the vast majority of cases his method could greately compress data and if it does he deserve credit for thinking out side of the box, I'd even guess to say in the non degenerate case outperform Huffman which is probably the "standard" lossless compression method as I think about 5 years ago I think I was taught that Huffman was provably the optimal lossless compression method.  The reason why he could outperform it is it will choose the best data arrangement while Huffman simply analyzes the existing data.

So yes.... his claim of reducing all files to 1 bit is ludicrus, I have never argued otherwise, but it is a compression method that IF it can be coded properly then it could be used recursively to shrink sizes of data recursively probably not to the extent he had hoped but maybe to a "pretty good" extend.  Commonly when I use WinZip or something I get what 1/2 compression, what if his method in the average case does 10X, 100X, 1000X?  A terabyte = 1 GB?  It is the blind ignoring of this POSSIBILITY that I disagree with simply because he made a single impossible claim or being able to reduce everything to nothing that you ignored EVERYTHING he said. 

Oh... and thanks for saying that other compression methods use tables [kinda sarcasm].  Where did I say all did?  I was simply stating why it wasn't "zero" data.  In his method it is possible for an entier string to be endoded into a single bit if the entire contents of a file just happens to be the most probably arrangements of bits.

"OK, so let's say that Einstein is behind the curve, let's say that all of the smart pirates now use torrents rather than p2p programmes."

SHHH.

Torrents?  YAWN.  People don't have to use torrents anymore (at least for movies).  They can watch live streaming movies from websites, who embed rather subtle advertising that scrolls across the top of the screen during the movie, and so they can make a profit and provide excellent bandwidth.  WHEN will the mainstream media THUGS figure this out and start innovating. 

BTW, if you can't afford ivy-league university, do not despair.  You can D/L pretty much any lecture on any topic from any of them.  Again, live-streamed, so you don't even have to have the files cluttering up your hard drive, and don't have to bother going through and deleting the files you've watched / listened to.

Perhaps MIT could actually INNOVATE something for a change and offer live streaming of university video lectures with embedded advertising from their OWN website.  But, alas, the simplest solution to the most complex problems are always the last to be considered by the so called 'best and brightest'.

I swear, all too often it seems there are only TWITS out there in the world who are supposedly running things.

Psyco,
and I could say read my first reply:D.  I said zero was impossible, I was agruing that his idea of rearranging to bits to increase your probability of "good" compression is a good idea which to my [very limited] knowledge of compression has not bee investigated AND if properly implemented could/would outperform other static lossless compression methods as it not merely analyzes the existing data but rearranges it.

Selur,
And unless you really like bringing raising the dead then drop the argument.  We agree and have never disagreed that yes you can't compress something to nothing in a lossless manner.  BTW your example is incorrect, 255 needs 8 bits to be represented 0-254 [255 unique numbers] is
00000000
to
11111110
you intended to say 7 bits.  But again, yes HE WAS INCORRECT TO CLAIM "ZERO" COMPRESSION I have never disagreed with that but again you all prefered to focus on his ignorance/over excited mistake and again ignore the possibility of exceeding other lossless methods simply because it is more fun/ easier to make fun of him.

Ok...? But I'm not certain what you are responding to.

oh, jolly. flintiepoo and his megalomaniac plans again. does someone take these seriously or is everyone just laughing inwards out of sheer politeness?

Actually any random data is compressible 1 time.... the issue is which system do you use?

What the smeg does this have to do with ending file sharing?

office is hardly a compression engine...

If on every server in the world there was a magic dictionary system of 100 gigabytes and they transmitted a mere 50 megabytes instead of 1 gigabyte you are saying due to the size of the dictionary it is not compression? Even if it applies to all files ever developed?