Hi Ken, thanks for this info. Are you saying that by
outputting a single
character rather than the text " " that the processor is
non-conformant
(I think you're saying that section 16.2 makes this optional), or that
getting an 0xFF byte in the stream rather than an 0xA0 is the problem?
It certainly would be non-conformant to output 0xFF here, but I suspect Ken
was jumping too quickly to blame the software. I would want to look very
carefully at how you invoked the processor - in particular, how the result
tree was serialized - and at exactly what you did to the output before
finding that 0xFF byte.
Michael Kay