Reliable Asynchronous Transfer Protocol (RATP)
RFC - Historic
(October 1984; No errata)
||RFC Editor Note
RFC 916 (Historic)
||Send notices to
Network Working Group G. Finn
Request for Comments: 916 ISI
RELIABLE ASYNCHRONOUS TRANSFER PROTOCOL (RATP)
Status of This Memo
This RFC suggests a proposed protocol for the ARPA-Internet
community, and requests discussion and suggestions for improvements.
Distribution of this memo is unlimited.
This paper proposes and specifies a protocol which allows two
programs to reliably communicate over a communication link. It
ensures that the data entering one end of the link if received
arrives at the other end intact and unaltered. The protocol, named
RATP, is designed to operate over a full duplex point-to-point
connection. It contains some features which tailor it to the RS-232
links now in common use.
We are witnessing today an explosive growth in the small or personal
computer market. Such inexpensive computers are not normally
connected to a computer network. They are most likely stand-alone
devices. But virtually all of them have an RS-232 interface. They
also usually have a modem. This allows them to communicate over the
telephone with any other similarly equipped computer.
The telephone system is a pervasive network, but one of the
characteristics of the telephone system is the unpredictable quality
of the circuit. The standard telephone circuit is designed for voice
communication and not data communication. Voice communication
tolerates a much higher degree of 'noise' than does a data circuit,
so a voice circuit is tolerant of a much higher level of noise than
is a data circuit. Thus it is not uncommon for a byte of data
transferred over a telephone circuit to have noise inserted. For the
same reason it is also not uncommon to have spurious data bytes added
to the data stream.
The need for a method of reliably transferring data over an RS-232
point-to-point link has become severe. As the number of powerful
personal computers grows, the need for them to communicate with one
another grows as well. The new markets and new services that these
computers will eventually allow their users to access will rely
heavily upon the telephone system. Services like electronic mail,
electronic banking, ordering merchandise from home with a personal
computer, etc. As the information revolution proceeds data itself
will become a commodity. All require accuracy of the data sent or
Finn [Page 1]
RFC 916 October 1984
Reliable Asynchronous Transfer Protocol
1. Philosopy of Design
Many tradeoffs were made in designing this protocol. Decisions were
made by above all ensuring reliability and then by favoring
simplicity of implementation. It is hoped that this protocol is
simple enough to be implemented not only by small computers but also
by stand alone devices incorporating microcomputers which accept
commands over RS-232 lines. Sophisticated but unnecessary features
such as dynamic window management [TCP 81] were left out for
simplicity's sake. Having several packets outstanding at a time was
eliminated for the same reason, and data queued to send when a
connection is closed remotely is discarded. This eliminates two
states from the protocol implementation.
The reader may ask why define this protocol at all, there are after
all already RS-232 transport protocols in use. This is true but some
lack one or more features vitally important or are too complex. See
Appendix II for a brief survey.
- A protocol which can only transfer data in one direction is
unable to use a single RS-232 link for a full-duplex connection.
As such it cannot act as a bridge between most computer
networks. Also it is not capable of supporting any applications
requiring the two-way exchange of data. In particular it is not
a platform suitable for the creation of most higher level
applications. Unidirectional flow of data is sufficient for a
weak implementation of file transfer but insufficient for remote
terminal service, transaction oriented processing, etc.
- Some of the existing RS-232 transport protocols allow the use of
only fixed size packets or do not allow the receiver to place a
limit on the sender's packets. Where that block size is too
large for the receiving end concentrator, that concentrator is
likely to immediately invoke flow control. This results in many
dropped and damaged packets. The receiver must be able to
inform the sender at connection initiation what is the maximum
packet size it is prepared to receive.
Show full document text