🚀 go-pugleaf

In article <21698@yunexus.YorkU.CA> oz@yunexus.yorku.ca (Ozan Yigit) writes:
> That is nice. I think you should publish the algorithm (and a through
> analysis of it) in some respectable journal.

I'm working on a paper.

> Coincidentally, this month's DDJ contains some stuff on this topic,
> including a compressor by Mark R. Nelson (model-2) that is 25-30% better
> than compress.

But how fast is it? How much memory does it use? Take, for example, the
important case of people transferring news between IBM PCs with size-64K
memory segments. compress works with 12 bits, or about a 6K input block
size on average. yabba works with a 12K input block size. whap works
with a 20K input block size. This alone gives a big improvement. Can
Nelson's modeller be used in limited memory?

---Dan