[ih] NIC, InterNIC, and Modelling Administration

[Matthias Bärwolff]

According to my research ca. one year ago, there are a host of prior
papers essentially anticipating in full the point made in the 1977
Davidson et al paper.

- RFC 137 (April 1971) sports an explicit reference to three protocol
  levels (and so do various RFCs and other NWG documents from the time)

- McKenzie (Host/Host Protocol for the  ARPA Network, NIC 8246, January
  1972) offers a good deal of elaboration on the higher layer stuff

- McKenzie (Host/Host Protocol Design Considerations, INWG General Note
  16, 1973) discusses some of the trade-offs in layering elegance
  versus the cost of passing (that is, copying) messages up and down

- Walden (Host-to-Host Protocols, 1975; reprinted in a 1978
  McQuillan/Cerf tutorial published by IEEE) comes up with the famous
  Figure with all the Arpanet protocols stacked on top of each other
  (and some bypassing others) -- that Figure also features in the 1977
  Davidson et al paper

However, the difference between protocols and interfaces was only
beginning to be appreciated then, which is readily apparent e.g. from
the Crocker et al 1972 paper (Function-Oriented Protocols for the ARPA
Computer Network, AFIPS 1972 Spring) (the one I take it Vint was
refering to below) which takes IMP-IMP protocol to be below the IMP-Host
protocol which in turn is below the Host-Host protocol.

Of course, IMP-IMP and IMP-Host are at the /same/ layer, upon which
there is sender-IMP-to-receiver-IMP (with the error correction and flow
control at that level), and then there is Host-Host, and then all the
app level stuff. This take on layering has been well put in Dave Clark's
1974 PhD thesis, but can also be infered fairly straightforwardly from
the BBN reports at the time, particularly the evolution of the ever more
distant host stuff that was added virtually right away in the very early
1970s.

Matthias

P.S. The above list is by absolutely no means to be taken as exhaustive,
especially in the INWG line of documents there are bound to be useful
elaborations of the layering notion, e.g. by Pouzin who in 1973 was very
adamant about clean and inviolable separation of layers already.

On Fri, Feb 18, 2011 at 11:32:28AM -0500, Vint Cerf wrote:
> i think there was a much earlier layering paper, lead author postel, 1972?
> Spring Joint or Fall Joint Computer Conference.

> v


> On Fri, Feb 18, 2011 at 11:02 AM, Craig Partridge <craig at aland.bbn.com>wrote:



> > > On 2/18/2011 6:45 AM, Craig Partridge wrote:
> > > > If you read the original OS layering paper by Hubert Zimmerman it is
> > > > clearly a top-down management work plan.  Useful to compare it with
> > > > the ARPANET layering paper of a few years later.  The difference is
> > Zim's
> > > > "here's how we'll break up the problem of developing standards" vs.
> > > > "here's why creating TELNET led us to a layered architecture".


> > > As I recall, documentation of the Arpanet approach to layering occurred
> > as a
> > > response to the OSI papers.  Prior to that it was de facto but not
> > documented

> > That was the received wisdom I got in 1983.  But when I went digging in
> > 1987 or so, I discovered it wasn't true.

> > The ARPANET layering paper was published by Davidson et al at the 5th
> > IEEE Data Comm Conference in 1977.  Zim's paper appeared in 1980.

> > Thanks!

> > Craig


-- 
Matthias Bärwolff
www.bärwolff.de