Hello, IID, you wrote: IID> And compressed by normal algorithms - though it is a little, but . Since arithmetics closely costs to theoretical to a limit for compression, And paternal there is a confidence, what ? Arithmetics is a method of entropy encoding. And that compression worked before it there should be a unit predicting probabilities. And this unit knows a lot of very important about the nature of the initial input data. That is in practice it select a miscellaneous at coding, for example, audio, the text or the machine code. But what character the data before enciphering/compression had it is not known, and without it it is necessary to use only the general methods of a prediction like PPM. But they simply will not have length of already oblate data that will be trained to predict something acceptable. Well that is even if to go on this way arithmetical coding here as though appears superfluous: After all simply enough to launch PPM on sequence and to look at the found regularities (if they are - the sequence means more likely it is not ciphered, and on the contrary). And to launch compression through arithmetics it is not so necessary. IID> and to it nobody presses, because of it . Well in itself arithmetics not so slow. It is quite comparable with the same classical coding of Haffmana. It is unpopular more likely for the historical reasons: for example because of problems with patents in some countries.