We're trying to port code from bash to ksh and we run in the following error:

export IFS=""
tempzip=`unzip -p file.zip file | uuencode -`
echo $tempzip | uudecode -

The system is a solaris 8 server. The end result is an error in ksh:

Invalid character (0x...)

In bash it works.

The end results have different byte counts depending on the shell in which the code is executed.

locale's are the same, both "C"

Don't use backticks and echo; those are the ones which will mess you up, not the actual uuencode. Both of those steps involve some amount of whitespace normalization and parsing of the strings you are handling there, whereas straight uuencode - | uudecode - would not involve any string handling by the shell itself.

As it turns out the shell was interpreting escape characters.