[Flac-dev] Blocking and compression.

[Wayde Milas]

On Fri, 2004-01-23 at 14:56, Josh Coalson wrote:
> this patent does not cover the FLAC reference encoder.  the
> entropy coder does not accumulate occurrences of individual
> symbol values, as specified in claim 1; it uses a single
> measure (mean or variance) of the overall residual distribution.
> 
> Josh

It could be argued that the mean and variance is computed from the
occurrences of individual symbol values; the mean and variance are a way
to store those computed values. I mean, how did you figure out the mean
and variance of the residual distribution? You looked at the individual
occurences of symbols and then came up with a way to represent them. The
accumulated values are represented in the mean and variance.

The whole thing is silly though. Hufffman computes symbol values.
Huffman stores the computed probabilities at the front of the stream.
(well, the first version did.. now adays the symbol probabilities are
recomputed on the fly after each symbol for speed/storage) Huffman has
been around since 1952. A whole slew of coders exist that do the exact
same thing. Gzip and Bzip2 being 2 of them.