[ih] Quantifying OSI
Michael Grant
mgrant at grant.org
Mon May 11 16:33:31 PDT 2026
Hey The Event, I was heavily involved in that! I worked at COS then. I
build and ran the testing lab at COS where all the vendors came in and
set up their stuff. I also did a lot of conformance testing of OSI back
then. Spent a lot of time in front of data scopes reading tcpdump like
traces to figure out why things couldn't communicate. It was often an
addressing issue.
To learn the protocol stack, I made myself a set of protocol cheat
sheets that looked like the ascii diagrams in the RFCs. I gave copies
to a bunch of people but was eventually told I couldn't distribute them
because they contained potentially copyrighted material from the CCITT
books. Not sure I still have those files around any more.
We, the tech folks, wanted, needed badly to get on the internet back
then. We ended up first mostly communicating using uucp to some of the
vendors others, it was fax. The management forbade us to get a ppp
dialup line for several years. It was laughable that we used uucp and
couldn't get x.400 working and I tried hard. Marshall Rose gave me a
set of ISODE tapes which I spent days and days trying to get something
talking to something outside of COS. If I recall, the issue was not
exactly ISODE but everything else around it. There was simply no OSI
Internet. You couldn't just get an X.25 connection and be on the OSI
net. Nobody sold CLNP but you could do it over ethernet. Nobody in
that time period ever imagined a telco dropping in a 10mbps ethernet
port and giving you pure CLNP, anything outside your own premise was to
be X.25 or maybe ISDN. On the application side, I don't recall, I don't
think ISODE X.400 talked to sendmail, it may have, i just can't
remember. Sun's X.400 definitely talked to sendmail though and from
memory, I had ISODE X.400 talking to Sun's X.400 but it was very much in
a lab playground. There was no authority to assign you a real globally
unique X.400 mail address, we simply made it up. There was no way to
route a message. There literally were no hostnames in OSI like in
TCP/IP!
Then there were no less than 5 different variants of transport named TP0
to TP4. Basically you ran TP0 over X.25 and TP4 over CLNP. Nobody
could tell me what happened if you had one end on TP0 and another on
TP4! Somehow X.500 was going to save the day and there would be some
sort of application layer gateway somewhere! Literally we had a
building (well, a couple floors of a building) filled with some of the
most clued up people on OSI and I talked to everyone and nobody had good
answers for the most basic questions of how this stuff was really
supposed to work. It was just a job. They were just doing stuff
because they were told to by higher ups. Nobody I worked with at COS at
the technical level believed this stuff was actually going to go
prime-time and everyone would be using it.
Anyway, yeah, The Event, that was a load of fun. That was possibly the
only time in the history of OSI that there was a small diverse group of
vendors equipment that actually talked to each other. After it
disbanded, everyone went home. There was never any way even anyone at
COS could use any of the stuff to communicate with anyone. We couldn't
even send an x.400 message to anyone after that. After The Event, it
was pretty much down hill. I talked to people internally at COS to see
if we could do something like Interop but there was no money, no desire
to do it. We sold a bunch of these super expensive OSI testers that ran
on Sun hardware and basically even if you passed the test suite, you
still wouldn't be able to talk to another implementation without a
stupid amount of config such that there really was no way ever to build
an internet with this stuff. Then the vendors started pulling out of
the consortium and then layoffs started.
I left COS for Sun and worked on federal bid & proposals for a couple
years where we saw the check-box requirement over and over from the US
Gov't and it was my job to tick those boxes with Sun's OSI
implementation. There, I met the folks at Sun France and they hired me
away from Sun Fed and I moved to Grenoble France where I could have
worked on Sun's OSI stack but by then, it was gathering dust and I ended
up working on SNMP and LDAP. I recall that Sun didn't even make its own
OSI stack, much of it was outsourced to some French company. I looked
at it a bit but the source was fairly impenetrable. Nothing was well
integrated into SunOS. Very messy. It was quite clear there was no
internal commitment to do this right.
Anyway, from start to finish of this episode of my life, I could never
understand how this stuff would really work and boy oh boy did I try to
make it work. Nobody around me either gave a crap if it worked. Nobody
around me at the technical level believed this stuff would overtake
tcp/ip which was getting more and more popular and more and more
companies were getting hooked in. Once you had a ppp connection, a
domain name, and started getting e-mail (if you hadn't already been
getting mail over uucp), you had not one iota of need or desire to get
x.400 working, nor anything else in that stack for that matter. X.500
morphed into LDAP and that morphed into Microsoft Active Directory
though LDAP still exists though I don't really know why people use it.
I had wondered for a long time why LDAP hadn't become some sort of
Internet like telephone directory. It would have been a nice place to
store things like PGP keys in but it never really took off on a global
scale. No one ever published their contact details publicly in LDAP.
Clearly this would have been a spammers dream to be able to just look up
everyone's address.
Aside from LDAP, I have run across things using ASN.1. I don't off hand
know if anything more than those subsets of OSI are still in use today.
------ Original Message ------
>From "Tom Lyon via Internet-history" <internet-history at elists.isoc.org>
To "Brian E Carpenter" <brian.e.carpenter at gmail.com>
Cc internet-history at elists.isoc.org
Date 11/05/2026 22:43:10
Subject Re: [ih] Quantifying OSI
>Anyone else remember the Enterprise Network Event - 1988 in Baltimore?
>It was the peak of hype for the MAP/TOP flavor of OSI, complete with IEEE
>802.4.
>Sun announced such a product there. Don't know if we ever sold any, but it
>got us past the dreaded corporate check-lists.
>
>https://historyofcomputercommunications.info/section/14.10/Enterprise-Network-Event-(OSI)-June/
>
>On Mon, May 11, 2026 at 2:22 PM Brian E Carpenter via Internet-history <
>internet-history at elists.isoc.org> wrote:
>
>> Yes, there was *enormous* expenditure by major companies worldwide that
>> thought OSI was the key to the future. IBM had people working on SNA/OSI
>> integration (including marketing vapourware), both in the US (Research
>> Triangle Park) and Europe (La Gaude). DEC (mainly at Littleton, MA)
>> invested many millions in DECNET Phase V. Boeing also spent millions, I'm
>> sure; I have no knowledge about GM. Untold numbers of companies spent both
>> marketing and development millions under the influence of US-GOSIP,
>> UK-GOSIP, European Commission policy, etc. Not to mention startups who
>> thought OSI was an enormous future opportunity. I imagine that this was
>> largely limited to North America, Western Europe and Japan, but it
>> certainly included a lot more than sending people to meetings. I have no
>> idea how to estimate the total but I suspect the correct unit is probably
>> the gigadollar.
>>
>> The emerging national research and education networks in Europe also spent
>> large fractions of their budgets on OSI preparedness - probably much less
>> money than industry was spending, but real enough, between about 1985 and
>> the early 1990s. The same went for NASA and DoE in the US.
>>
>> Standards goers and their fine lunches and dinners were probably quite a
>> small fraction of the real total cost.
>>
>> Regards/Ngā mihi
>> Brian Carpenter
>>
>> On 12-May-26 08:43, Karl Auerbach via Internet-history wrote:
>> > Our company (Epilogue Tech.) was involved with ISO/OSI mostly via the
>> > MAP and TOP efforts by General Motors and Boeing.
>> >
>> > Those gatherings tended to be somewhat well attended, although I don't
>> > think many of the attendees were people who actually implemented things.
>> >
>> > --karl--
>> >
>> > On 5/11/26 3:46 AM, John Day via Internet-history wrote:
>> >> That would be hard to calculate for the US. There were 5 OSI main
>> committees each with 15 - 20 or more participants in various subgroups. In
>> the US, US corporations paid for the time and travel of their participants.
>> Some companies (IBM, Honeywell, ATT, etc) would have multiple participants
>> in the same committee. There were multiple US meetings between major
>> international meetings every 9 months and international sub-group meetings
>> between the 9 month major meetings.
>> >>
>> >> The cost was all paid by the companies participating. In addition,
>> there were 802 meetings that were feeding into the OSI work. This was
>> especially true of how network management was gotten off the dime to get
>> around IBM stonewalling. All other 802 standards were process by an ISO
>> committee, because some countries saw IEEE as a US organization.
>> >>
>> >> I wouldn’t even hazard a guess at how many people or companies were
>> participating from the US. The Europeans did complain sometimes abut the
>> large US delegations to the meetings.
>> >>
>> >> Take care,
>> >> John Day
>> >>
>> >>
>> >>
>> >>> On May 10, 2026, at 23:53, Carl Malamud via Internet-history <
>> internet-history at elists.isoc.org> wrote:
>> >>>
>> >>> Hi -
>> >>>
>> >>> I’m trying to quantify the size of the OSI efforts. I’ve seen the 25
>> >>> million ECU investment by the EU, and have some pointers to US
>> government
>> >>> efforts. Has anybody tried to collect these numbers?
>> >>>
>> >>> Also very interested in non-monetary indicators. I have easy access to
>> >>> number of IETF participants and count the traffic on mailing lists. Any
>> >>> similar metrics for OSI? The best indicator so far is “many fine
>> lunches
>> >>> and dinners” but surely there has to be something more scientific.
>> >>>
>> >>> With best regards,
>> >>>
>> >>> Carl
>> >>> --
>> >>> Internet-history mailing list
>> >>> Internet-history at elists.isoc.org
>> >>> https://elists.isoc.org/mailman/listinfo/internet-history
>> >>> -
>> >>> Unsubscribe:
>> https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>> --
>> Internet-history mailing list
>> Internet-history at elists.isoc.org
>> https://elists.isoc.org/mailman/listinfo/internet-history
>> -
>> Unsubscribe:
>> https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>>
>--
>Internet-history mailing list
>Internet-history at elists.isoc.org
>https://elists.isoc.org/mailman/listinfo/internet-history
>-
>Unsubscribe: https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>
More information about the Internet-history
mailing list