by Darius Kazemi, May 8 2019
RFC-128 is titled “Bytes” and is authored by Jon Postel of UCLA. It's dated April 21st, 1971.
The technical content
Postel discusses the issue of mismatching byte size between hosts. What if a sending host specifies a 5-bit byte and the receiving host specifies a 3-bit byte? He says that this could be considered an invalid connection and immediately terminated, or it could be considered valid and there would have to be some kind of translation that takes place.
Postel says that UCLA thinks asymmetry in byte sizes should be allowed and that the way they would be handled would be by essentially considering everything to be a bit stream. This removes the need for padding-based algorithms on the NCP side in the case where you insist that “your” byte is always a full byte and must be considered such by the receiver (though he doesn't rule out using this system as well).
Postel suggests that RAND's Data Reconfiguration Language might be used for padding-based case.
Handling things on the bit level seems to have the effect of pushing the byte-level translation work from the NCP to the end-users instead, which is probably a good idea. I think the NCP should be as simple as possible.
There are a couple papers on the Data Reconfiguration Service floating out there. This 1971 paper is co-authored by a bunch of ARPANET regulars is a kind of high level overview, and this 1972 paper by Harslem, Heafner, and Wisniewski is far more in-depth and essentially a technical manual including source code.