Re: Call for participation -- Is OSI really useful?

John Day <Day@bbn.com> Wed, 01 June 1994 14:00 UTC

Received: from ietf.nri.reston.va.us by IETF.CNRI.Reston.VA.US id aa01965; 1 Jun 94 10:00 EDT
Received: from CNRI.RESTON.VA.US by IETF.CNRI.Reston.VA.US id aa01961; 1 Jun 94 10:00 EDT
Received: from sun2.nsfnet-relay.ac.uk by CNRI.Reston.VA.US id aa15245; 1 Jun 94 10:00 EDT
Via: uk.ac.ulcc.vmsfe; Wed, 1 Jun 1994 14:29:33 +0100
Via: UK.AC.NSFNET-RELAY; Wed, 1 Jun 94 14:22 GMT
Received: from BBN.COM by sun3.nsfnet-relay.ac.uk with Internet SMTP id <sg.05282-0@sun3.nsfnet-relay.ac.uk>; Wed, 1 Jun 1994 14:21:43 +0100
Sender: ietf-archive-request@IETF.CNRI.Reston.VA.US
From: John Day <Day@bbn.com>
Subject: Re: Call for participation -- Is OSI really useful?
To: laurae@laurae.ar.telenex.com
Cc: 73543.1077@compuserve.com, agrawala@cs.umd.edu, Baker@forty2.enet.dec.com, baos@oss.com, bob@uci.com, colin <colin@intelsat>, conrad@oss.com, cpd@one.com, dallas@forty2.enet.dec.com, day@bbn.com, DBRITT@nctsemh-npt.navy.mil, dchoi@vax2.cstp.umkc.edu, devon!ed-kelly@mhs.attmail.com, dicksc@uci.com, dicksw@uci.com, dyons@arch4.att.com, eric@isci.com, ews@ctt.bellcore.com, frank@cos.com, gray@osi.ncsl.nist.gov, heather@tandem.com, jmhunt@atlsita.org, kk@arinc.com, kuiper@osison.osiware.bc.ca, lee@vax2.cstp.umkc.edu, lee@huachuca-jitcosi.army.mil, lee@ntd.comsat.com, lowe@osf.org, lrajchel@attmail.com, markh@rsvl.unisys.com, mkao@cup.hp.com, p.furniss@ulcc.ac.uk, quigley@cup.hp.com, rdesjardins@attmail.com, ron11@cc.bellcore.com, rschilk@huachuca-jitcosi.army.mil, sjg@arch4.att.com, THINOSI@ulcc.ac.uk, troisi_brenda@tandem.com, truoel@gmd.de, truskows@cisco.com, vantrees@sed.stel.com, wdavison@rlg.stanford.edu, X3T5@osf.org
In-Reply-To: <9405311333.AA00734@laurae.ar.telenex.com>
Date: Tue, 31 May 1994 22:02:45 -0500
Mail-System-Version: <BBN/MacEMail_v1.5@BBN.COM>
X-Orig-Sender: THINOSI-request@ulcc.ac.uk
X-ULCC-Sequence: 187
X-ULCC-Recipient: ietf-archive%us.va.reston.cnri@uk.ac.nsfnet-relay
Message-ID: <9406011000.aa15245@CNRI.Reston.VA.US>

Laura
> When I sat down earlier this week to define the agenda for the 
> upcoming "Birds of a Feather" on OSI Upper Layer efficiency, I 
> found myself sifting through so many completely varied solutions, 
> I forgot the problem being solved. And then I picked up an issue 
> of Network World (May 9) that a co-worker had left on my desk. 
> Thomas Nolle's column "Reality Check" that week was titled "OSI 
> decline reflects problems in standards process." So I read the 
> column. Nolle, in referring to OSI, claims that "the idea was to 
> give the marketplace a standard, and everybody would go along with 
> it from the start, ...thus was born the open systems movement, 
> and the OSI model." However, a more realistic reality check is 
> the recent history lesson I received from Bob Stover. He reminded 
> me that OSI was defined as a gateway and backplane architecture 
> to connect very large, private networks together for the purpose 
> of increasing market share for each party. There's a reason why 
> the Reference Model looks like a mutant cross between ARPANET 
> and SNA. Given that, twenty years after the first RM was created, 
> we now use terms like "collaborative computing" and "common data 
> repository" and "information infrastructure" instead of "gateway" 
> and "backplane" the question is, what problem are we trying to 
> solve? 

I am afraid Bob's view of things is colored by where he was at the time,
i.e. a mainframe maker.  For many of us who had been doing networks for
6 - 8 years; had been involved with the ARPANet, Cyclades, NPL or EIN;
had been working in IFIP WG6.1 developing international transport and
terminal protocols, had been living on e-mail for 5 years; this was the
next logical step to making it ubiquitous. (I had been living 1000 miles
from my job and commuting electronically for two years when OSI started.
In fact, I had been getting all of my development resources over the net
for 5 years.  You haven't lived until you develop an operating system
for a machine that is 2000 miles from development environment and
downloading over the net. )  There was no influence from IBM on the
fundamental nature of the RM.  In fact since SNA was a hierarchical
terminal concentration network, it was precisely what we were trying to
avoid.  The biggest influence and most detrimental was from CCITT.  At
this time, the PTTs, especially the French were trying to duplicate the
accomplishments of the networks mentioned above.  However their
tradition caused them to get a lot wrong, but was close enough that it
was difficult to say no.  This was especially true of the Europeans for
who the PTT legal and market clout was immense, nothing like it is
today.  Unfortunately, I must say that your interpretation of the early
view was more that of the late comers trying to catch up.  It had little
to do with the direction that those of us who had been living on
networks had in mind.

>The BOF was initiated to research efficiency options for 
> OSI upper layers. However, if the paradigm is no longer applicable, 
> and the problems facing the communications industry today are 
> not being addressed, how valid are the solutions we've produced? 
> More to the point, is OSI really useful?
> 
The fundamentals of the paradigm are still useful. The upper layers were
screwed up early, but I am now very confident that structure is "right"
and can grow cleanly and easily to what it needs to be.  Working out the
true nature of p-ctx was the last piece of that puzzle.  As for the
lower layers, it is only slightly off.  A small number of adjustments
and major things fall into place.  But then many of you saw that in my
talk at NIST a few weeks ago.  It turns out that multicast, mobility,
and a lot of other things end up being very straightforward.

> The essential point in Nolle's article is not the usefulness of 
> the RM, but rather on the effectiveness of the standards process. 
> He asks, "Who are these people writing standards anyway?" It is 
> a good question. We all know that attendance at meetings has 
> dropped off significantly. In addition, the standards bodies, 
> ANSI and ETSI, ISO and ITU-T, and the regional workshops, OIW, 
> EWOS, and AOW, are all legislative bodies. 

The reasons for standards are still the same whether in the IETF, ISO,
or ITU.  The question one must carefully weigh is which process is going
to give good solutions in the long term.  Are the processes sceptible to
block voting that is going to allow the process to be hijacked.  Notice
that the IETF can be hijacked by one or more companies sending lots of
people; ISO can be hijacked by regional block voting; ITU is no place
for you if you are not a phone company, etc.  Pick your poison.  The
alternative is to adopt as standard whoever sells the most.  This will
not be the technically superior solution.

>Therefore they are 
> not the optimal place to conduct scientific research and initial 
> technical work. Such groundbreaking work should be completed 
> before and outside of the legislative process. 

Research can not be done in standards committees.  Actually it could but
the vendors have made sure that it won't be.

>Vendor forums have 
> been sucessful at creating new technology, but can be restrictive 
> in their membership and cost prohibitive. All of the regulations 
> and fees for membership in any of these organizations is limiting 
> the participating community to just a handful of focused experts. 

Actually this does not have to be the case and hasn't been in Europe. 
In the US where we have had a very bizarre policy for funding basic
research, it has and will continue to be the case.

> Given the initial scope of OSI, that it was a gateway and backplane architecture
> for large private networks, this was exactly the way 
> it should be; but if we agree that the scope has changed, then 
> members of the user community and the academic community should be
> contributing more to the effort. What, if anything can we do to 
> expand the process to increase participation?
> 
Who cares about OSI?  The Internet Suite is older than the OSI protocols
and every one of their protocols is suffering from scaling problems. 
European political agendas dominated too many technical decisions in
OSI.  One can not expect the IETF to create a cohesive new group of
protocols.  What you want is not necessarily the OSI protocols, although
a few are salvageable:   TP4, CLNP needs a little work to simpify it,
A2CSE as the basis for applications, etc.  

For those of you familiar with Chicago politics, if the Internet is the
dirt road being paved as the NII, the contractor building it is putting
too much oatmeal in the concrete.  The potholes will be so bad, it will
need repaving before it is finished.

Take care
John