It sounds like you have an encoding problem. First suggestion would be is, find out what your personal encodign is, and regexp/tr away the invalid crap.

i.e. if you are using 7 bit ascii only, then killing off anything over 127 would at least clean things up a bit. Just make sure not to kill off the XML structuring symbols, such as #, (space), ",', >, < and the other chars by accident. Try egress filtering.. leaving only what you want in, in.

I feel your pain. tye's scratchpad, while it may work with some xml parsers, has char 0x001, or whatever char isn't liked by mine, per the options of the spec.. so i have to clean it up when parsing on my end.

Update: just testing something w/ katterbox - sporty's scratchpad