This also includes some incidental fixes around the description of the record
size at the beginning of the encoded message.
FYI @davidben and @nyaxt. I went with the 0-len->0-len option because decoders need to deal with content that's too short for the 8-byte record anyway, so I think the special-case fits fairly naturally there. Let me know if that turns out not to be true in the actual implementation.
This also includes some incidental fixes around the description of the record size at the beginning of the encoded message.
FYI @davidben and @nyaxt. I went with the 0-len->0-len option because decoders need to deal with content that's too short for the 8-byte record anyway, so I think the special-case fits fairly naturally there. Let me know if that turns out not to be true in the actual implementation.