From: Ethan Katz-Bassett (ethan@cs.washington.edu)
Date: Tue Oct 05 2004 - 23:12:04 PDT
This paper presents a protocol to interconnect packet switching networks.
It describes how these networks can be connected through Gateways and how
hosts on these networks can communicate using the described "transmission
control program" (TCP). These concepts seem to be new contributions of the
paper. The authors consider many aspects involved in interconnecting
distinct networks and present a (mostly) clear and coherent protocol. In
some sections (for instance, the discussion of how to format segments), the
authors clearly present alternative and discuss their reasons for favoring
one. I appreciated these explications. The paper explains TCP's
retransmission and duplicate detection mechanisms in detail, and they seem
versatile, powerful, and lightweight.
The authors acknowledge many of their works' limitations. Much of the work
desperately needs to be validated and made explicit by implementation and
experimentation. The paper's focus on high-level design keeps it
to-the-point, but experimental results showing, for instance, the
performance of the retransmission policy would strengthen it. The authors
leave for elsewhere details such as how to keep a receiver from accepting a
duplicate transmission received after terminating the original association.
I think that the paper does not adequately explain its final concept, in
which TCPs "allocate RCB/TCB pairs. which rendezvous with the exchanged data
and then disappear," leaving me unclear about this idea. Finally, the
benefit of hindsight shows problems, including the consideration of packet
charging policies and the assumption of the adequacy of 8-bit network
addresses. The current factoring into TCP/IP layers provides a separation
of concerns and allows the possibility of different protocols on top of IP;
the authors present their ideas in a limiting single layer.
The work remains relevant today. Most elements of the protocol remain in
use. The paper presents a solution to many of the difficulties in
internetworking and, presumably, served as a launch pad for later work.
This archive was generated by hypermail 2.1.6 : Tue Oct 05 2004 - 23:12:09 PDT