[e2e] Jon Crowcroft's thought questions...

Lyman Chapin lyman at acm.org
Mon Jul 28 14:22:37 PDT 2003


At 7:04 AM -0400 7/26/03, Craig Partridge wrote:
>In message <E19gIpL-0000AU-00 at wisbech.cl.cam.ac.uk>, Jon Crowcroft writes:
>
>>i) who was involved in the ISO
>>OSI Teransport Protocol (e.g. class 4 was quite nice -we did an 
>>implementation
>  > which
>>outperformed tcp:)
>
>I'd page Ross Callon for this one (he did CLNP and as I recall did a lot on
>TP4 as well).  Also John Burruss and Lyman Chapin comes to mind (perhaps
>incorrectly).

Craig and Jon,

Most of what became TP4 can be traced back to the Cyclades transport 
service developed by Louis Pouzin, Hubert Zimmerman, and others in 
the 1970s. At that time, the principal forum for the research 
community's discussion of TCP and Cyclades TS was IFIP WG6.1 (more 
familiarly "INWG"); these discussions led to the publication of INWG 
General Note #96, "Proposal For An Internetwork End to End Protocol" 
(V. Cerf, A. McKenzie, R. Scantlebury, and H. Zimmerman) in July of 
1975. [NB: The first version of what would become the landmark 
Cerf/Kahn 1974 paper "A protocol for packet network interconnection" 
was presented at an INWG meeting in Sussex in 1973; essentially, TCP 
and TP4 evolved from, if not a common ancestor, then certainly a 
common "primordial soup"!] When the OSI work started a few years 
later, INWG 96 - and several years' worth of practical experience 
with both TCP and the Cyclades TS - provided the starting point for 
TP4.

The INWG members in the late 70s and early 80s deserve most of the 
credit for TP4. After the INWG work was pulled into OSI, the list of 
contributors to the eventual TP4 standard includes, in addition to 
John Burruss and me, Jean-Pierre Ansart, Dick Watson, Hoyt Kesterson, 
Jim Moulton, Dave Piscitello, Dave Oran, John Day, Terry Knowles, 
Jeremy Tucker, and John Neumann - as well as the many people I'm sure 
I've inexcusably, but inevitably, neglected to mention.

Chapter 12 of the book that Dave Piscitello and I wrote about TCP/IP 
and OSI (available in PDF at 
http://www.interisle-group.com/OSN/OSN.html) compares TCP and TP4 at 
length, but makes no claims about one outperforming the other :-) The 
TP4 implementation I wrote for Data General in the mid-80s ran over 
CLNP in DG's proprietary AOS and AOS/VS operating systems, and we had 
TCP/IP only in DG's version of Unix (DG/UX), so I have no first-hand 
data to add to Jon's observations about the relative real-world 
performance of the two protocols.

- Lyman




More information about the end2end-interest mailing list