Presented by

  • Paul Wayper

    Paul Wayper

    Paul started writing code for a XZ-80 on a typewriter, and things have improved a bit since then. After a long career as a sysadmin and support engineer, e currently works for Red Hat trying to automate that so that other people don't have to. He writes mostly Python and his computing interests range from hardware and microprocessors to massively parallel distributed computing. In his spare time he turns wooden bowls, skis, plays music, tunes a 3D printer, and collects spare hobbies just in case.


Compression algorithms are used on everything from web traffic to satellites, from backups to real time. Most of the popular general compression algorithms are based on LZ77, which as its name implies was published in 1977 by Abraham Lempel and Jacob Ziv. But a year later they produced another algorithm, LZ78, which became the basis of LZW with Terry Welch's improvements in 1984. Despite being used most commonly in the GIF image format, it doesn't get much attention these days because it is not as efficient as the current techniques such as bzip2, lzma and the recent zstandard. But why? What does LZW get wrong, and can it be improved? In his talk, Paul will go into his research into improving the LZW algorithm, give a demonstration of encoding and decoding, and compare its compression ratios to other current compression methods. YouTube: