Definition of "character" position for oa:start and oa:end

Is there ambiguity in the definition of "character" for a Text Position
Selector
<http://www.openannotation.org/spec/core/20130208/specific.html#TextPositionSelector>
?

Are positions given as:

   - UTF code points,
   - UTF-8 code units,
   - UTF-16 code units or
   - some other definition of "character"?


Interoperability between implementations may suffer if the definition of
"character" is not stated explicitly.


I recently raised a similar issue
<https://github.com/IDPF/epub-revision/issues/555> (with some useful links
to other standards) against the epubcfi specification
<http://www.idpf.org/epub/linking/cfi/epub-cfi.html>. The resolution in
that case was "UTF-16 code units", which is unsurprising given the
web-technology nature of that specification.