0.0.9q uses a new command line for calling mplayer.  The "correct" command
line for mplayer seems to change with different mplayer versions.  


0.0.9p handles a recent change in mplayer command line arguments.  It switches
between the old format and the new format based on mplayer's response to the 
mplayer --help invocation.  This version also introduces a feature of checking
to see whether both files actually exist before printing that the match was
found.  This allows you to leave your ~/.fdmf database alone and let it grow.
Old deleted "ghost" files in the database will not be considered matches, but
these records are useful to keep because if an identical file is encountered 
later with a different name, the cached summary will be used, which is a huge 
speedup.  This feature can be disabled in vector_pairs.h for special uses.

0.0.9o (unreleased stupid experiment)

0.0.9n uses mplayer by default for decoding music files.  The reason for this
is that mplayer is faster than mpg123 for decoding mp3 files to raw pcm.  
Also, fdmf will no longer exit if it encounters a directory that it cannot
read.  It just prints a warning and continues.  README and INSTALL have
been updated.  Supported file formats (via mplayer) are mp3, ogg, m4a, wav,
wma, and realaudio.  Basically, whatever mplayer will play will work in fdmf.

0.0.9m is the first fdmf version that tries to completely avoid using the
shell.  Passing arguments to external programs through the shell had been
causing a lot of quoting/escaping problems. 

0.0.9l fixes several string handling bugs.  fdmf no longer
calls the openssl program.  It uses Digest::MD5 to compute hashes.

0.0.9j adds a check to the fdmf perl script to see if the MD5
hash of the data of the file that we are currently processing is the same as
the MD5 hash of the data of another file that we have already indexed.  If
it is, we now skip running sonic_reduce() and instead simply reuse the 
summary vector that is in the database for the other file.
