There's currently a thread on the IRTF end2end list about the history of well-known numbers for network ports (such as 80 which is the standard HTTP TCP port). I did some checking in the older RFCs and discovered the earliest mention of ports was in RFC 33 (February 1970), in which they were called sockets. As time went on, people developed services which were assigned ports in somewhat of an ad hoc manner. It wasn't until RFC 349 (May 1972) that Jon Postel, who went on to head the IANA, proposed that he be the individual who maintained and handed out the assignments for well-known sockets. The ARPANET NCP protocol was in use at this time, but the concept was carried over into contemporary TCP and UDP (as ports).
It was interesting reading some of the oldest RFCs. This is the first time I'd really spent serious time with them. You can see the origins not just of the protocols, but of principles and conventions that exist to this day. One I found particularly interesting was RFC 615, a proposal for a network standard pathname, which resembles somewhat (in concept) today's URI. Even back then (March 1974), there was a sense of what the future would look like.