RFC 167 (rfc167) - Page 2 of 4

Socket conventions reconsidered

RFC 167                                  Socket Conventions Reconsidered


The current NCP Protocol says nothing about how hosts should assign
socket numbers to process ports, except that the low-order bit is to
specify socket gender (i.e., send or receive). Two recent proposals call
for additional network-wide conventions on the 32-bit socket-number. The
first proposal asks that a portion of the socket number be reserved for
a network-unique user number for accounting and access control. The
second proposal asks that the high-order 16 bits of the socket number be
zero to assist smaller hosts in reducing the space required for socket
number tables.

It is recommended that both of these proposals be set aside.  Because a
large perturbation of the current NCP Protocol is required to provide
adequate handles for accounting and access control, and because the
socket number is already underpowered for its use, it is recommended
that both proposals be set aside until serious consideration can be
given to a major NCP Protocol overhaul.

DISCUSSION

The socket number, as it is used in the current NCP Protocol is a small
number with a big function. It will probably be found that a
substantially more powerful identification mechanism (e.g., a
hierarchical naming scheme with arbitrarily long names) is required to
satisfactorily manipulate process ports. Two features of such a
mechanism will be (1) that it treats accounting and access control with
the respect they deserve, and (2) that it is part of a simpler NCP
Protocol more easily implemented under the existing size and complexity
restrictions of smaller hosts.

Socket numbers are process port identifiers used in establishing
connections between processes. It is essential that they be UNIQUE to
avoid ambiguity during connection. It is important that their assignment
to specific processes be REPEATABLE for reconnection on a regular basis.

To assure that process port identifiers are unique and repeatable it is
necessary to subject their allocation to access controls.  The simplest
of access controls assuring uniqueness is that provided by NCPs which
check their tables of active connections for duplication when a process
requests the use of a specific socket number.

There is real difficulty in constructing schemes for allowing socket
number assignments to be repeatable. Some socket numbers are to be
universally known and associated with processes operating with specified
protocols (e.g., a logger socket, an RJB socket, a file transfer
socket). Other socket numbers might not be universally known, but given
to their users in a transmission over a universally known socket (e.g.,
the socket pair specified by the transmission over the logger socket
using the Initial Connection Protocol (ICP)).  Concurrently running