> I remember showing PhDs about PKZIP and having them not believe it was possible to compress data without losing information - I had to literally show them the (rough) algorithm.
They were right; that is a well-known and trivial-to-prove theorem about lossless compression.
Lossless compression works because you only apply it to very particular types of data. It doesn't and cannot work on general data; that's why we have specialized compression algorithms for every different kind of data.
I took graduate compression in the mid-90s.. JPEG was already invented and we studied that.. there was some handwaving but it was not as primitive as some comments make it out to be today
of course -- JPEG includes means to adapt to the image data patterns, however, which is mentioned in these threads as architecturally important to useful lossless compression, too. (did you really think that 'JPEG is lossy' is more than a 'gotcha' reply?)
They were right; that is a well-known and trivial-to-prove theorem about lossless compression.
Lossless compression works because you only apply it to very particular types of data. It doesn't and cannot work on general data; that's why we have specialized compression algorithms for every different kind of data.