LZW was originally designed for implementation by special hardware, but it

turned out to be highly suitable for efficient software implementations too. An enhanced

variant is available on UNIX systems and many other systems as the compress command.

We refer to this variant as LZC. The speed and compression performance of LZC are a

result of careful data structure design for hash table look-ups, some tuning and the addition

of logic for restarting the algorithm when the source file changes its characteristics

enough to worsen compression performance.