1. Aack! It won't compile! It doesn't work! What do I do?
2. Why do yabba and whap use so much memory?
3. Why does squeeze do better than yabba and whap on long files?
4. What does ``block size'' mean?
5. Why doesn't yabba take a filename argument?
6. When will there be a new version of yabbawhap?


1. Aack! It won't compile! It doesn't work! What do I do?

Read through README, Makefile, and QUESTIONS. Make checkconf, run it,
and read carefully through its output. Then try again. If you still have
trouble, send the author a note at brnstnd@nyu.edu.


2. Why do yabba and whap use so much memory?

With -UPTRS -DTYPE=short you will typically use half as much memory on a
32-bit machine. With -DHASHTYPE=short you will save some more memory for
yabba and unyabba, though not for whap or unwhap. These changes usually
cost a bit of speed. See Makefile for more details.

You can also change the block size by changing NODEMAX and MOD. These
don't really affect speed, but they do affect compression. With them you
can make yabba and whap use as much or as little memory as you like.


3. Why does squeeze do better than yabba and whap on long files?

By default, yabba and whap are compiled to use similar amounts of memory
to compress. squeeze uses a lot more memory, so it can keep track of
more patterns in the input. If you compile yabba and whap with NODEMAX
and MOD so high that they use as much memory, they'll do better than
squeeze.

For example, on a 97338-byte shell archive, yabba -m65533 compresses to
39266 bytes, and whap -m65533 compresses to 37358 bytes. yabba -m100000
compresses to 36673 bytes, and whap -m100000 compresses to 33566 bytes.
(In contrast, compress produces 43322 bytes, squeeze produces 33943
bytes, and lharc produces 32455 bytes. A Huffman coder layered on top of
yabba or whap would produce even better results.)

There are occasional files where squeeze performs better than yabba and
whap no matter how much memory you use, but never by much.


4. What does ``block size'' mean?

The number after -m represents a number of input characters. -m100000
means that yabba (or whap) will try to find patterns in a stretch of
input up to 100000 characters long. NODEMAX and NODENUM are on the same
scale. Here 100000 is the block size. In the above example with a
97338-byte file, any -m value above 97338 will produce the same results.

(For compress, the natural block size is the number of patterns it is
keeping track of at once, not the number of input characters. With -b14,
compress will keep track of 2^14 or 16384 patterns at once. Typically
16384 patterns would stretch over 60000 input characters. -b16 is
roughly equivalent to -m300000.)


5. Why doesn't yabba take a filename argument?

Wait for the -f option. Separate programs, fyabba and funyabba, will
apply yabba and unyabba to files; fyabbadir and funyabbadir will work on
directory hierarchies.


6. When will there be a new version of yabbawhap?

I am actively working on the next release; I have targeted version 1.50
with most of the planned extensions for May 1. Send any suggestions or
patches to me at brnstnd@nyu.edu.