Roger Moser wrote:
(as Wayne pointed out to me) there still could be a
slash, for example in the macro "%{d./}".
Sigh.
So choosing "%!" instead of "%/" did no help.
It helps a bit, because the delimiters can only appear
"within" macro-expand in the context of a domain-spec.
But if it's messy whatever you do then the obvious %/
is probably better.
If unknown-mechanism is deleted, then unknown-modifier
is the only place using macro-string. This could be
simplified to...
unknown-modifier = name "=" *( macro-literal / "/" / "%" )
...because nobody wants to parse an unknown-modifier.
The next space or EOF terminates it. After deleting
all macro-string it can be used for a "macro-with-sp":
macro-string = *( macro-literal / macro-expand / "/" / " " )
For the next draft we need signed confirmations by all
participants, that this is _really_ the final syntax.
Or Wayne's "validating evaluation", that's like burning
all bridges you have crossed.
Bye, Frank