[ih] protocol structure -> economic structure

[John Day]

I don't think one could say "protocol" does and I don't see how a 
mapping that preserves some invariance would have any effect on it, 
but architecture certainly does and has been used for this and well 
understood for decades.

Since the mid-70s, we have talked about how the PTTs used the 
"beads-on-a-string" model to define market boundaries.  It is hard to 
say who first pointed this out.  But by 1975, Pouzin was giving talks 
along this line and creating quite a stir.  I remember a session at a 
conference in Toronto in probably 77, when Louis was doing one of 
these talks (complete with slides of the User hanging from a gallows 
on the parapet of the PTT castle) ;-), the session was running over 
time and people were collecting at the edges of this very large room. 
The session chair stopped Louis to say we were over time and Louis 
said what would you like me to do? And the *audience* en masse said 
"GO ON!, GO ON!"  ;-)  Meanwhile the PTTs people in the front row 
were furiously taking notes.  I don't think I have seen an audience 
react like that before or since!

It was this criticism of the PTTs that got CYCLADES shut down and 
Louis essentially black-listed in France for 30 years.

I started drawing pictures and using them in lectures back then to 
illustrate it.  It was amusing that one could take the same picture 
of a terminal connected to a small host connected to switches 
connected to a host could be labeled asymmetrically for the X.25/PAD 
environment the PTTs were pushing and the same picture could be 
labeled symmetrically to illustrate terminals connected to TIPs and 
hosts on the ARPANET. (I think I used this in an article in a special 
issue of Computer in the late 70s or early 80s.)  The whole reason 
for difference in labeling and the beads-on-a-string model in general 
was to define who got to sell what.

In my talks on this subject, I often point out that the impact of the 
CYCLADES model, i.e. a layered model with a reliable end-to-end 
transport over a datagram network, relegated the PTTs to a commodity 
business.  They couldn't claim that the "value-added services" on 
hosts belong "in the network."  Both catch phrases of the day. (It 
has been amusing to watch the more gullible audience today swallow 
the PTT argument dressed up as "cloud computing." Different players, 
same game.)

This of course was the primary cause of the connection/connectionless 
debate (that and the shift from a deterministic model to a stochastic 
one, but money always speaks louder).  But I highly doubt this 
originates with me. Similarly, I point out that IBM's SNA was a 
hierarchical architecture and the emerging new layered model of peers 
were incompatible.  One can always take a peer architecture and 
subset it to be hierarchical (and we have, e.g. client server), but 
it is impossible to take a hierarchical architecture and make a peer 
architecture.

I discuss some of this in my book (Chap 3 and 10), but this far from 
the first nor original with me.  I learned to watch the money from 
Pouzin and others!  ;-)

There are ways that protocols do effect the market too.  But 
normally, that is more tactical than strategic.  Usually in the form 
of someone has an implementation or a product and getting their 
approach or something close to it accepted as the standard gives them 
an advantage.  The IETF's requirement for two implementations plays 
to this game very nicely.  Although I am sure there are counter 
examples.

I really doubt that anyone was first. Anyone in business knew this 
and was acting on it.  (Or they didn't stay in business long.)  The 
technical arguments and their proponents were merely pawns and 
ammunition tin the battle over where the market share went.

There is much more one could say but I have rambled enough.

Take care,
John Day

At 12:10 +0000 2012/03/02, Tony Finch wrote:
>There's an idea that the structure of a protocol (or the topology implied
>by the protocol or that emerges from it) determines the economic structure
>of the service providers and users of that protocol. Sadly I can't now
>remember where I got this idea from.
>
>Does anyone here have pointers to literature mentioning or describing this
>idea? Is anyone acknowledged as its originator? Does it have a snappy name?
>
>Tony.
>--
>f.anthony.n.finch  <dot at dotat.at>  http://dotat.at/
>Forth, Tyne, Dogger, Fisher, German Bight: Variable 3 or 4, becoming south or
>southeast 5 or 6, occasionally 7, perhaps gale 8 later in Fisher. Slight or
>moderate, occasionally rough later in Dogger, Fisher and German Bight. Fog
>patches, occasional rain later. Moderate or good, occasionally very poor.