Steve Lawrence from the
Daffodil project has observed that we're not particularly consistent
about delimiters and the %ES; entity.
We allow WSP* with caveat that you
can't have initiatedContent, but we disallow ES.
There is a use case for optional delimiters e.g.,
initator="%CR;%LF; %ES;"
meaning absorb and discard an optional CRLF.
Without %ES we end up having to model this syntax as an optional hidden element.