🚀 go-pugleaf

We'll ignore the technical merits of ISO, because in the end it is
marketing forces that seem to determine success and failure of
standards more than their merit.  The overhead of ISO is probably
twice that of TCP/IP for many purposes, but if vendors are willing to
put enough resources into making ISO fly, this probably will not be
critical for most users.

It looks like systems vendors are tending to move towards ISO for
networking, but are hedging their bets with TCP/IP.  TCP/IP has a big
head start, is still more widely available, and has public domain
software.  I think it's going to be aheard for the forseeable future
in universities and among hackers.  I think the first real splash for
ISO is going to be general availability of DECnet phase V.  However
from what I can tell of it, it's sort of a cross between traditional
DECnet and ISO, and some services (particularly virtual terminal) may
not interoperate with non-DEC ISO stacks. (Can anyone confirm or deny
that?)

I suspect within a few years the way it's going to end up is that
vendors who now have proprietary protocols are going to tend to
replace them by ISO, but will also have TCP/IP software available, and
vendors that currently use TCP/IP will probably continue to use mostly
that, but will also have ISO available (but usually at a significantly
higher cost).  My intuition is that there's more networking going on
using proprietary protocols now than TCP/IP, so if that all migrates
to ISO, ISO will end up ahead, particularly in shops that get their
networking from their system vendors rather than having active
networking staffs of their own.  Of course major routers and other
networking equipment will support both TCP/IP and ISO.

The research community will probably continue to use TCP/IP, for
several reasons:

  - the process of defining and modifying the protocol is
	easier for researchers to deal with.  
	Protocol documents are easy to get, and there are
	reasonably low-overhead ways to put out new protocols
	and extensions, particularly for experimental purposes.
	In general TCP/IP is user-driven and ISO is
	vendor-driven, not because the ISO folks are in any
	sense closed, but because of the economics of
	participating in the standards community.  The TCP/IP
	process is intentionally designed to allow participation
	via computer mail and other low-cost methods.

  - implementations are widely available and generally
	inexpensive.  It seems that most ISO implementations,
	except those that come bundled with systems are
	still expensive.

  - the protocols are enough simpler that it's easier to do
	experimental work with it.

I've wondered if ISO is going to be a replay of PL/I or Ada, which are
situations where a large company or the government tried to impose
standards, but generally failed.  However in all fairness I think
there's wide enough support among vendors for ISO that it will become
widespread.  The timing is hard to guess.  The past 5 years have all
been the year when ISO was finally going to take off.  I still think
DECnet is going to catalyze it.  Unfortunately DECnet phase V has been
delayed incredibly, but it looks like it will finally be out in 1991.

One thing that might change my view is if the system vendors that now
offer TCP/IP decide to offer ISO at the same cost (including bundling
it free with systems that have TCP/IP bundled free).  Perhaps Berkeley
4.4 will catalyze this the same way 4.2 catalyzed widespread
availability of TCP/IP.