Recall that the contents of fo:block are defined as
(#PCDATA|%inline;|%block;)*
and that there are some interesting complications in the contents of "&inline;" which, frankly, I still don't understand in spite of clarifications from the editors.
One complication is whitespace because of the unfortunate mix of block level FO and text. There's also the unpleasant problem that <fo:block> <fo:marker... may unexpectedly trigger a fatal error because of the whitespace before the fo:marker. Yeah, unintended consequences of seemingly easy and straightforward definitions. Doh.
J.Pietschmann
