Showing papers on "Communications protocol published in 1975"

Communication protocols and error recovery procedures

Abstract: The design of multiple computer systems and computer networks poses some interesting new problems and shows some older problems in the light of a new context. We are interested in the problems which are related to the design of communication protocols and error recovery procedures. In the design of computer systems, error recovery procedures have often not received very much attention, although they are essential for the reliable operation of most systems. In the case of multiple computer systems, appropriate error recovery procedures can be used to obtain a very much increased availability of the system services. This is also an advantage of computer networks. Whereas single computer systems can sometimes function to a certain extent without elaborate error recovery procedures, data communication protocols, on the other hand, normally have to contain explicit error recovery procedures, since the underlying communication line is normally not very reliable. Therefore, approaches that have been used for the design of protocols in computer networks could also be useful in the design of local multiple computer systems for obtaining higher reliability and availability.

Method of data communications in a heterogenous environment

Abstract: A method of data communications for use in systems employing serial transmission techniques in a ring or loop configuration. Access to the communication network is via a communications processor and a interface unit termed a node. The communications processor is responsible for network protocol and activities as well as for interfacing devices to the network. The node provides the only connection to the loop and drives the clock for the communications processor as well as the node itself from the transmitted data. All information in the communication network is transmitted in a modified Mauchly format and will appear as n-bit data or control characters. A binary ONE is represented as a signal level transition from one level to another, while a binary ZERO is represented by an opposite signal transition. A control character, which always precedes a data character, will either be the address of a node or a null. To read information from the network, the node must detect a control character which represents its address. To insert information into the network the node must detect a control character which represents a null.

Communication protocols in a network context

Abstract: In a distributed computers network based on a packet-swltched communication subnet, it is possible to describe the system as a hierarchical structure. We will consider here only three levels:-level 0: communication between nodes of the subnets.-level 1: communication between hosts connected to the net.-level 2: communication between users processes (subscribers) in different hosts.At each level, the communication is based on a protocol and the data structures to be exchanged are different. For instance the data structure exchanged at level 2 may be a sequential file. As there is no direct support of communication at this level, it is through the mechanisms of level 1 and 0 that the transfer will take place with data structure at each level not directly related to the upper level one. The complete definition of the system will therefore required not only the level 0, 1 and 2 protocols, but also inter-level protocols.In the following, we will concentrate on the level 1 protocol. It is the basic communication protocol of a network since it will be used by user processes in different hosts and it will use the subnet as a communication support. This level 1 entity is called a “TS” (transport station) in CYCLADES [13] and a “TCP” (transmission control program] in [2].

An integrated approach to network protocols

Abstract: Host-to-host protocols (H-H) for heterogeneous computer networks are still in infancy. So far very few implementations are in existence. Among those on which documentation is available are Arpanet and Cyclades. The former provides only for basic services allowing the transfer of up to 1000 octet messages, with flow control but not error control. The latter allows up to 32 000 octet messages, with error and flow control. Both are similar in the sense that they offer only a message transfer service, which is intended for building higher level protocols more appropriate for specific uses. Since data to be transferred are usually structured in various ways, a traditional approach is to superimpose additional layers of specific protocols, each one dealing with a particular level of structure. While being functionally correct, this approach leads to heterogeneity, redundancy and overhead among the various layers.

Commentary on procedure calling as a network protocol

Abstract: Issues in designing distributed computing systems. Shortcomings of RFC 674; see also RFCs 542 and 354.

ARPA Internetwork Protocols Project Status Report

Abstract: : An Internetwork Communication Protocol was implemented which permits reliable interprocess communication to be established between hosts in different networks. Estimates of program size are made for a single connection version of the program for use in a mobile packet radio terminal. Preliminary throughput measurements indicate serious problems with the Very Distant Host interface and its related software. A plan for detailed analysis of delays in the system has been made and the results will be reported during the next quarter.