Network Working Group J. Postel
Request for Comments: 902 J. Reynolds
ISI
July 1984
ARPA-Internet Protocol Policy
Status of this Memo
This memo is a policy statement on how protocols become official
standards for the ARPA-Internet and the DARPA research community.
This is an official policy statement of the ICCB and the DARPA.
Distribution of this memo is unlimited.
Introduction
The purpose of this memo is to explain how protocol standards are
adopted for the ARPA-Internet and the DARPA research community.
There are three important aspects to be discussed: the process, the
authority, and the complex relationship between the DARPA community
and the DDN community. To do this some background must be given and
some of the players described.
DARPA = Defense Advanced Research Projects Agency
DDN = Defense Data Network
The DARPA World
The DARPA world is headed up by the DARPA office. DARPA sponsors
research on many subjects with a number of contractors. This set of
contractors is called the "DARPA research community". DARPA
typically casts its research efforts into "programs" that involve
work by several contractors. One program is the "Internet Program",
which is researching computer communications issues and constructing
experimental communication systems. When the experiments are
successful, the results are often put into use to support further
work in the Internet Program and other DARPA research programs. In
this way, DARPA developed the ARPANET, SATNET, Packet Radio Networks,
and the Internet.
In 1981 DARPA established the Internet Configuration Control Board
(ICCB) to help manage the DARPA Internet Program.
Postel & Reynolds [Page 1]
RFC 902 July 1984
DARPA Internet Protocol Policy
The ICCB
The concerns of the ICCB fall into two categories:
Short Term Issues:
Keeping the Internet operating as an on-going resource, i.e.,
dealing with problems that arise due to the growth in the size
of the system and the level of use of the system. Sometimes
this suggests research on new procedures and algorithms, or
suggests changes to the existing protocols and procedures.
Sometimes the results of long range research become available
and their introduction into the current system becomes a short
term concern.
Long Term Issues:
The ICCB also considers communication problems related to the
Internet more abstractly. The ICCB suggests to DARPA possible
research topics and experiments. The ICCB may act as a
sounding board for ideas suggested by others.
DARPA has delegated some aspects of the management of the Internet
Program and operation of the (experimental) ARPA-Internet for the
DARPA research community to the ICCB.
The members of ICCB were chosen to represent a spectrum of interests
and viewpoints. The ICCB members are from different organizations,
their individual backgrounds specialize in different operating
systems and their viewpoints on computer communication issues are
diverse.
The chairman of the ICCB is also the "Internet Architect", and the
assistant chairman is the "Deputy Internet Architect". The ICCB
currently has 12 members. The Internet Architect is Dave Clark of
MIT, and the Deputy Internet Architect is Jon Postel of ISI.
The DDN World
The DDN is a communication system for DoD operational use. It
integrates many networks and communication systems now used and
planned within the DoD. One part of the DDN system is networks that
are also part of the Internet, specifically MILNET and the networks
connected to it.
The DDN is managed by the DDN Program Management Office (DDN-PMO).
The DDN-PMO sets policy for the use of DDN facilities and enforces
protocol standards established for use in the DDN networks.
Postel & Reynolds [Page 2]
RFC 902 July 1984
DARPA Internet Protocol Policy
Within the DoD, there are three protocol committees: the Protocol
Standard Steering Group (PSSG), the Protocol Standard Technical Panel
(PSTP), and the Protocol Configuration Control Board (PCCB). These
committees have members that represent most elements of the DoD.
Generally, they develop and recommend guidelines for protocol
standardization and usage to the DDN-PMO, and to all of the DoD. The
PSSG is a policy setting committee for all of DoD on matters of
protocols standards.
The Relationship between the DARPA World and the DDN World
There is cooperation between DDN-PMO and DARPA about the Internet. A
few people serve on both the DoD committees (PSSG, PSTP, or PCCB) and
the DARPA committee (ICCB). There are good working relationships
between the key people in the DARPA office and the DDN-PMO, and
between the technical people in both worlds at lower levels.
For example, the ICCB may decide that a certain protocol is to be
used in the ARPA-Internet, and develop an implementation plan and
schedule. The DDN-PMO would separately consider the issue. It may
decide to require that protocol to be implemented in DDN on the same
schedule, or it may decide to wait for some results from the DARPA
experiment with that protocol before committing to a schedule, or it
may decide that that protocol is not required in the DDN.
There are two documents that specify TCP. RFC-793 is the official
specification of the DARPA research community. Military Standard
1778 is the official specification of the DDN community. The two
documents specify the same protocol.
Organizations that are connected to the Internet through authority
derived from DARPA follow the rules set by the ICCB and DARPA.
Organizations that are connected to the Internet through authority
derived from DDN-PMO follow the rules set by the DDN-PMO.
DARPA Official Protocol Designation
Official protocols for the ARPA-Internet and DARPA research community
are specified in RFCs and should have that designation indicated in
the first few paragraphs of the defining RFC. That is, the RFC
defining an official protocol should have a policy statement that
says,
"This RFC specifies a standard for the DARPA community. Hosts on
the ARPA-Internet are expected to adopt and implement this
standard.",
Postel & Reynolds [Page 3]
RFC 902 July 1984
DARPA Internet Protocol Policy
or something quite similar.
Also, there is a memo titled "Official Protocols". This document is
issued occasionally as an RFC that describes all the official
protocols of the ARPA-Internet. This document provides information
on each protocol; its status (experimental, required, etc.),
specification, additional comments, other references, dependencies,
and the person to contact. The most recent issue is RFC-901.
RFCs are coordinated by the RFC Editor and distributed by the Network
Information Center (NIC). The RFC documents are stored as online
files in the NIC's computer. Announcements of new RFCs are sent to a
mailing list of interested people. The RFC Editor is Jon Postel.
The Normal Development of an DARPA Protocol Standard
There probably never has been a "normal" case. In most instances
some exception or another has been made to the following procedure.
The Typical Chain of Events
The development of a protocol starts with some discussion with
random people in messages and meetings over an idea of a new
protocol and the form it ought to take.
Someone writes a draft and proposes this draft to a group of
people who are interested in the problem. They suggest revisions
and iterate the discussion. Eventually, they may decide that they
have a reasonable definition of the new protocol and then pass
this definition on to the RFC Editor.
The next step is that the RFC Editor sends a draft to other people
who might also be interested in the problem. These people can
number just a few, or be part of a large mailing list. Depending
upon the results from this selected informal group, the draft can
be revised and rewritten several times.
When this process stabilizes, the protocol draft is sent out as an
RFC, identified as a draft proposal of a protocol that may become
an official protocol. The RFC is sent to the ARPA-Internet world
at large.
After a certain amount of time, if only a few comments are sent
back, some people may try to implement the draft protocol.
Test implementation of a protocol is a difficult management issue.
Postel & Reynolds [Page 4]
RFC 902 July 1984
DARPA Internet Protocol Policy
Experiments must be done with a small number of participants due
to the difficulty in changing many implementations at the same
time if changes in the protocol are necessary.
In cases where the ICCB deems it necessary, a set of test
implementations will be done. A few participants are picked
(typically 5 or fewer) for such experiments. This may lead to
revision of the protocol before further implementations are
encouraged or before the protocol is made official.
If no problems arise, a new RFC is issued containing the complete
definition of the protocol, and that the protocol is an official
protocol of the ARPA-Internet and DARPA research community.
In general, lower level protocols are more critically judged than
higher level protocols (for example, a protocol like TCP would be
subject to more careful study than an application like the DAYTIME
protocol).
The Bottom Line
For the ARPA-Internet and the DARPA research community, DARPA is in
charge. DARPA delegates the authority for protocol standards to the
ICCB. The ICCB delegates the actual administration of the protocol
standards to the Deputy Internet Architect.
For the DoD in general, the PSSG is in charge. The PSSG delegates
the authority for the day to day management of protocol standards in
the DDN to the DDN-PMO.
Postel & Reynolds [Page 5]