Version 1.06.0 (5/11)

Compression.

Substantial reworking of compression to accommodate
BZIP2 compression.   The Apache Bzip library is used or
since this is very slow, the user can specify a local
command to do the decompression using the BZIP_DECOMPRESSOR
environment variable.  This is assumed to require a
'-' argument which is added if not supplied by the user.
The decompressor should act as a filter between standard input
and output.

User compression flags are now completely
ignored and the compression and the compression
is determined entirely by the content of the stream.
The Apache library will be needed in the
classpath to accommodate BZIP2 inputs if the user
does not supply the BZIP_DECOMPRESSOR.

Adding additional compression methods should be much easier and
may only involve adding a couple of lines in the
FitsUtil.decompress function if a decompressor class
is available.

One subtle consequence of how compression is now handled
is that there is no advantage for users to
create their own BufferedDataInputStream's.
Users should just provide a standard input stream
and allow the FITS library to wrap it in a
BufferedDataInputStream.

Other changes.

A bug in the UndefinedData class was detected
Vincenzo Forchi and has been corrected.

The nom.tam.util.AsciiFuncs class now handles
ASCII encoding to more cleanly separate this
functionality from the FITS library and to enable
Java 1.5 compatibitity.   (Suggested by changes of L.Bourges)
Other V1.5 incompatiblities removed.

The HeaderCommentsMap class is now provided to enable
users to control the comments that are generated in system
generated header cards.  The map is initialized to values
that should be the same as the current defaults.  This
should allow users to emulate the comments of other packages.

All Java code has been passed through NetBeans formatter
so that it should have a more uniform appearance.