Erron writes:
Maybe in a version 5 sig., simply make the signature subpacket length
definition packet equal to a two octet scalar - this would then marry with
the total length of all un/hashed subpackets and would also reduce the
amount of processing required to determine the length of a signature
subpacket.
Maybe so. We have had many complaints that the current variable-length
packet-length mechanism is somewhat over-engineered for the purpose.
Still, almost all subpackets are short enough that one byte is enough
for the length, so you do save a few bytes by using the compressed
length encoding.
Hal