From: Erika Rice (erice@cs.washington.edu)
Date: Tue Oct 05 2004 - 22:53:28 PDT
Erika Rice's review for Vinton Cerf and Robert Kahn's "A Protocol for
Packet Network Intercommunication":
The paper "A Protocol for Packet Network Intercommunication" by Vinton
Cerf and
Robert Kahn can, judging from both its age and its outlining of ideas
fundamental to the architecture of the Internet be considered one of the
classics in the domain of network design. The paper contains a fairly
detailed
description of a packet switched internetwork protocol. Although they call
this protocol a transmission control protocol (TCP) it differs from
modern TCP
in that it also considers addressing concerns, which are now factored
into IP.
The paper "A Protocol for Packet Network Intercommunication" by Vinton
Cerf and
Robert Kahn can, judging from both its age and its outlining of ideas
fundamental to the architecture of the Internet be considered one of the
classics in the domain of network design. The paper contains a fairly
detailed
description of a packet switched internetwork protocol. Although they call
this protocol a transmission control protocol (TCP) it differs from
modern TCP
in that it also considers addressing concerns, which are now factored
into IP.
The paper gives a detailed description of some of the technical problems
related to creating a protocol that can work with many different packet
switching networks. It is consistent with Clark's paper "The Design
Philosophy
of the DARPA Internet Protocols" in that it rejects solutions which do now
allow diverse networks to connect to the internetwork.
The TCP described in the paper is similar, at least at the fairly high level
this paper aims to describe, to the modern TCP/IP combination. As such,
it was
a good, although certainly pedagogical description of the modern Internet.
The paper gives a detailed description of some of the technical problems
related to creating a protocol that can work with many different packet
switching networks. It is consistent with Clark's paper "The Design
Philosophy
of the DARPA Internet Protocols" in that it rejects solutions which do now
allow diverse networks to connect to the internetwork.
The main failings I see in this paper are, in some ways, failures to see the
future. Thus, while failings, it is really impossible to completely
blame the
authors for these short comings. For example, the justification for the
size
of the addressing space seems quaintly naive to modern readers.
Similarly, the
idea that congestion, although planned for, would be rare seems rather
optimistic. Altogether, it serves as a warning that one should carefully
consider the basis for assumptions that limit designs.
Of course, one must also, somewhat facetiously, consider the fact that using
footnotes in mathematical expressions is always a dubious practice. What
should have been a perfectly logical statement that the window size must be
less than half of the number of sequence numbers was unnecessarily
confused by
footnote number three, making it appear that the window size must be smaller
than the strange value of 1/(2^3) or 1/8 of the number of sequence numbers.
Cerf and Kahn give a valuable insight into the architecture of internetworks
and the Internet in particular. Their paper also helps one to
understand the
evolution of even the most basic aspects of the Internet architecture.
It is
instructive to realize TCP/IP did not sprung fully grown from the minds of
researchers but evolved as they saw what was necessary and extraneous, what
failed and what worked.
This archive was generated by hypermail 2.1.6 : Tue Oct 05 2004 - 22:53:30 PDT