The decoding routine decodes %xx into the equivalent character.  In the case of
%2b, the resulting character is the plus sign (+).  Separately, if the decoded
or undecoded character is a plus sign, it is replaced with a space.  I think
that the space for plus replacement should not take place if a decoding just
took place.  In a modification we made, we replaced this:

if (b == '+') {
  b = (byte)' ';
}

with this:

else if (b == '+') {
  b = (byte)' ';
}

That way, the just-decoded plus sign won't be turned into a space, but a literal
plus sign will be turned into a space.  As far as I can tell, this is the
correct behavior, and corrects a problem we were having storing files with plus
signs in their file names in Slide repositories.

We are using a modified version of SLIDE_2_1_M1_RELEASE on various platforms,
but the code quoted above is still present in the most recent CVS versions.