The non-normative backwards compatibility appendix in F+O (Appendix D) lists some incompatibilities in string functions that no longer exist: they were removed during the last call period. Since this appendix is informative only, I suspect there is no objection to us correcting the information during PR.

Note also that in most of these cases, the function signature given in the appendix doesn't match that in the body text.

I haven't done a complete trawl, but the incorrect statements include:

concat(): If compatibility mode is off, an error will occur if ... argument is a number or a boolean.

starts-with(): In 1.0, returns false if the first argument is an empty node-set. In 2.0, returns ()

contains(): In 1.0, returns false if the first argument is an empty node-set. In 2.0, returns ()

substring-before(): In 1.0, returns "" if the first argument is an empty node-set. In 2.0, returns ()

substring-after(): In 1.0, returns "" if the first argument is an empty node-set. In 2.0, returns ()

substring(): In 1.0, returns "" if the first argument is an empty node-set. In 2.0, returns ().

normalize-space(): In 1.0, returns "" if the first argument is an empty node-set. In 2.0, returns ()