Summary

  • RFC 787 circulated A. Lyman Chapin's case for connectionless data transmission as a companion to connection-oriented service in the OSI Reference Model. A self-contained unit could be sent without first establishing a connection at that service layer.
  • The July 1981 cover memo asked for comments before a possible ANSI X3T5 review and later ISO presentation. The working paper itself announced that it had yet to receive committee approval or become a US consensus position.

Some communications are worth opening a long conversation for. Others are a thermometer saying it is still warm. A. Lyman Chapin used that difference to challenge the assumptions of the emerging OSI Reference Model. His RFC 787, circulated in July 1981, carried a working paper arguing that connectionless data transmission deserved a place beside connection-oriented transfer. A datagram could set off with one self-contained piece of information. The proposal about datagrams faced a longer itinerary: collect comments, go to ANSI, perhaps travel to ISO. Standards work had its own handshakes.

The cover memo set out the route with unusual clarity. US organizations interested in international data-communication standards would review the paper through September 1, 1981. Chapin would revise it using their comments. The ANSI committee handling OSI Reference Model questions, ANSC X3T5, might then consider it. Approval there could make it the basis of a US position to the relevant ISO groups. On the paper itself sat a conspicuous warning that it had not been reviewed or approved by the proper technical committee and did not yet represent US consensus. The RFC archive gave readers the argument; the committee had not supplied the stamp.

Chapin's technical complaint was about the shape of the model. A connection-oriented service builds an association before moving data and can attach sequencing, recovery and other state to it. That can be excellent when a long exchange needs those properties. A connectionless service, as the paper described it, moves a single self-contained unit from source to destination without prior negotiation at that service layer and without its ordinary assurance of delivery. Higher-layer users remained free to add timers, acknowledgments, retransmissions or sequence numbers.

The architecture could offer the simple unit without forcing every application to borrow a connection's entire wardrobe.

The examples were the paper's best argument. A large set of sensors might send periodic measurements to one control point. If one fresh reading vanished, the next could arrive soon; the application would notice a prolonged absence. A network could also announce a changed server address or a shutdown warning to many recipients. A one-question, one-answer service could treat the answer as the useful acknowledgment. Chapin pointed to point-of-sale, reservations, inventory and banking systems as examples of this interaction, and to power-system control centers where measurements, requests and broad commands shared one operating environment.

He also discussed group addressing and time-sensitive voice or telemetry, where waiting for an old sample might be worse than using the next one. These were cases for choosing a service shape, not a promise that one service would suit every conversation.

The document even resisted an easy marketing shortcut. A service that confirms delivery or nondelivery of each datagram had the surface shape of one operation, but Chapin called it a hybrid carrying assurance from connection-oriented behavior. The distinction matters because “connectionless” was a statement about the service's own obligations. An application might build reliability above it; that did not turn the underlying one-unit service into a long-lived connection. The question for the OSI architects was where those obligations belonged.

RFC 787 organized its case as a survey and tutorial for people who had to fit protocols into layers. It asked for connectionless service to run alongside connection-oriented service, with each used where its costs and behavior made sense. Its status was equally specific: Chapin had sent a working paper to reviewers and described a conditional path through ANSI and ISO. The document did not record a final committee decision. Its 1981 significance lies in the effort to bring working patterns such as self-contained measurements and one-shot requests into the architecture being debated.

The packet in this story took a short route. The idea about the packet took a long one, because a standard changes what many implementers will be expected to build. Chapin's paper gave the committee concrete work to examine: a thermometer report, a broadcast notice, a single request, each with a reason to travel before a connection could be negotiated. The handshakes in the meeting room remained necessary. The ones on every tiny message were up for debate.