Charles Lindsey wrote:
Thanks for the pointer to RFC 2978. I see that '%' is also allowed, but
not '*'. So writing a foolproof parser for extended-initial-value would be
tricky (you have to scan from R to L until you have identified the two
rightmost "'"s), but I doubt implementors will bother to go so far.
That is not the only way to skin that particular cat; a grammar-based
parser can easily handle extended parameters including a hypothetical
charset name containing a single quote and/or percent sign without
resorting to reverse scanning -- it's handled by the normal parser
state machine stack. It's not even particularly "tricky" -- one
simply writes the appropriate grammar rules and it just works.