Canonical Huffman Coder Construction
The first time I ever heard the phrase Canonical Huffman Coder was in reference to the technique used in PKZip to store Huffman tables. I don’t know where the technique originated, but it is basically a way to construct a Huffman table so that the actual codes don’t have to be stored when storing the table. This makes for some nice space savings when compared to a first-pass naïve implementation. (Like the ones I’ve done in the past.)
It turns out that somebody named Gareth was attempting to implement this code but was having a bit of trouble. His post to comp.compression brought out some useful help from some of the newsgroup regulars, and did a lot to shed light on the topic, and includes a reference to the paper that actually gave birth to the concept.
