Snobs Miss the Point
From Zanecorpwiki
While I was tarring a file, I remember thinking back to my college CS times and remembering how people loved to point out that there is no reversible compression algorithm that can make all possible input smaller. The proof is pretty simple.[notes 1] At best, you can make some things bigger and some things smaller and it ends up being a zero sum game. In practice, compression always yields a slight increase in size over the entire universe of input because of encoding overhead.
But that's a silly point to make. The fact is that inputs are not random. Sure, loss-less compression of a series of random images of digital static will almost always end up larger than the input, but who cares? There's no need to compress bits. You might as well just generate new random bits next time you need random bits. It's random. It doesn't matter what it was.[notes 2]
It's easy to be a snob because stuff is so contextual. Snobbery is about false rigor, and poking holes in things that violate such rigor. Don't get me wrong, rigor is vital and laziness is a bigger problem than snobbery. The weird thing about snobbery is that it still takes a lot of energy... like, it's no harder to be a connoisseur instead of a snob, and you're always better off. At least, that's the way it seems to me.
Notes
- ↑ First, consider that the input and output from the program is the same stuff. Usually a sequence of bits, but the idea holds however you model the input and output. If the game is to try and make every input smaller, you soon see it becomes impossible because once you make one string smaller, then that displaces another small string.
- ↑ If it's very random at least.
