More on selecting sequence numbers

  • Authors:
  • Yogen K. Dalal

  • Affiliations:
  • Digital Systems Laboratory, Stanford University, Stanford, CA

  • Venue:
  • Proceedings of the 1975 ACM SIGCOMM/SIGOPS workshop on Interprocess communications
  • Year:
  • 1975

Quantified Score

Hi-index 0.00

Visualization

Abstract

This note serves to supplement and extend the ideas and issues raised in [TO74]. Tomlinson examines the problem of establishing a connection (association), and being able to detect delayed packets of an old incarnation of a connection, when it is being opened and closed in quick succesion, or when the connection breaks owing to some crash and is restarted later. No protocol can guarantee reliability of communication given certain types of crashes [SU74], [BE74]. This last issue is out of the scope of this note. First the problem of establishing and closing a connection is further examined and the claim that a three way exchange is sufficient for single message transfer in [TO74] (even under conditions of a correctly functioning system) is refuted. Next the mechanism for implementing the “clock-driven” initial sequence number choice is examined in detail showing how various parameters interact, and the algorithms neccessary in order that it work. Finally alternative schemes to replace “resynchronization” are considered and the tradeoffs shown.