Yes, let's just stick with ABNF.
>>> Randy Turner <rturner at sharplabs.com> 06/06 3:44 PM >>>
> If we are mixing binary and ASCII values within our encoding ...
See my earlier posting measurements we did of our prototype where
we encoded both ways
"(binary length)name(binary length)value"
and then just
Not a lot of difference in our timing measurements. In fact it takes more
time to encode using binary lengths.
Doesn't seem worth it to me to skip a great legacy and ease extensibility
reduce risk of interoperability for almost no processing gain on a
portion that is only very minimally involved in the overall end-to-end
time required from application-to-paper.