Alex van den Bogaerdt wrote:
Your examples (below) are, indeed, other cases where the spec
doesn't work.
That's not the case, the spec. is unambiguous for valid input.
If you'd also want "unambiguous syntax errors" it gets messy,
What's wrong with "neither <directive> nor <modifier> => bad
<term>" ? The ABNF is not about counting the minimal changes
to transform an invalid input into a valid policy. Check out
RfC 2822 or 3986 (STD 66) for similar constructs. Sometimes
it's even okay to have ambiguous matches (only as long as the
semantics is identical or irrelevant of course). Bye, Frank