Conversation with Bob Kahn
Notes from a 1969 discussion about character code conversion and Host-IMP communication.
Crocker, S.D., "Conversation with Bob Kahn", RFC 6, DOI 10.17487/RFC0006, April 1969, <https://www.rfc-editor.org/info/rfc6>.
@misc{rfc6,
series = {Request for Comments},
number = 6,
howpublished = {RFC 6},
publisher = {RFC Editor},
doi = {10.17487/RFC0006},
url = {https://www.rfc-editor.org/info/rfc6},
author = {S.D. Crocker},
title = {{Conversation with Bob Kahn}},
pagetotal = 1,
year = 1969,
month = apr,
} RFC text
Network Working Note Steve Crocker, UCLA
RFC-6 10 April 1969
CONVERSATION WITH BOB KAHN
I talked with Bob Kahn at BB&N yesterday. We talked about code conversion
in the IMP's, IMP-HOST communication, and HOST software.
BB&N is prepared to convert 6, 7, 8, or 9 bit character codes into 8-bit
ASCII for transmission and convert again upon assembly at the destination
IMP. BB&N plans a one for one conversion scheme with tables unique to the
HOST. I suggested that places with 6-bit codes may also want case shifting.
Bob said this may result in overflow if too many case shifts are necessary.
I suggested that this is rare and we could probably live with an overflow
indication instead of a guarantee.
With respect to HOST-IMP communication, we now have a five bit link field
and a bit to indicate conversion. Also possible is a 2-bit conversion
indicator, one for converting before sending and one for converting after.
This would allow another handle for checking or controlling the system.
The HOST can send messages or portions of a message to its IMP specifying
1. Tracing
2. Conversion
3. Whether message is for destination IMP or HOST
4. Send RFNM
5. HOST up or down
6. Synchronization
7. Format Error Messages
8. Master Link Clear
9. Status Requested
The IMP can send to its HOST information on
1. Conversion
2. REFNM Arrived
3. IMP up or down
4. Synchornization
5. Called HOST not Responding
6. Format Error
7. Status in IMP
I also summarized for Bob the contents of Network Notes l, 2, and 3.
AI ELI5
On this page
§1Character Code Conversion
The ARPANET connected hosts with different word sizes and character encodings. PDP-10 computers used 5 seven-bit ASCII characters packed into 36-bit words; IBM 360s used EBCDIC codes in 32-bit words. The conversation with Kahn discusses BBN's plan to convert all data to and from the standard NVT-ASCII format in the IMPs, with character code tables specific to each host. Crocker suggests that overflow from case shifting might be acceptable since it is rare.
Some computers think words are 6-bit puzzles, others think they're 8-bit. The network agreement says 'everyone translate your puzzle to the standard shape, send it through, then translate back'.
§2Host-IMP Protocol Details
The note documents Host-to-IMP communication fields and signals: a 5-bit link field to identify the connection, a conversion bit, and status information. Hosts can send commands about tracing, conversion control, message destination, RFNMs (ready-for-next-message), host up/down status, synchronization, error format, and link clearing. IMPs respond with status on conversion, RFNM arrivals, IMP status, synchronization, host responsiveness, format errors, and internal state.
A host tells its IMP what to do (trace this, convert that, is the host ready?) and the IMP tells back what happened (code converted, message ready, I am working).
§3Historical Note
This RFC is an informal working note, not a specification. It captures early ARPANET design discussions that informed later protocol work. It shows the practical concerns engineers faced when building the first packet-switched network connecting heterogeneous computers.
This is not a rule book, just notes someone wrote down while working on a hard problem. It shows what people cared about in 1969.