[ih] OSI and alternate reality

[Jack Haverty]

On 3/15/24 03:19, David Sitman via Internet-history wrote:
> In my talk at the EARN 40th Anniversary Conference in Athens in April I
> would like to speculate a bit about what the world would be like today if
> OSI had won the "Protocol Wars".
> In 1986, it was a foregone conclusion that EARN would migrate to OSI in the
> near future. However, when I began my international activity in 1991, OSI
> was discussed as a promise that had gone largely unfulfilled and EARN
> members were actively supporting TCP/IP networks. It seemed obvious why
> TCP/IP had prevailed.
> Would we have seen the same rapid and universal adoption of computer
> networking with OSI? Could the Web have flourished? Would address space and
> security issues be alleviated? Would "OSI on Everything" have become a meme?
> I would be very grateful for any thoughts about this.
>
> Thanks,
> David Sitman

Well, I've always liked sci-fi "alternate reality" stories.  So here's 
some thoughts...

IMHO, both the TCP/IP and OSI approaches were similarly incomplete back 
in the 80s and 90s when the marketplace was choosing the Internet 
approach.   Evidence for that is the deluge of 1000s of RFCs since then, 
containing Standards, Draft Standards, Wannabee Standards, Protocols, 
Algorithms, and other such technical additions that have been deemed 
necessary over the decades and continue today.

Both TCP and OSI technologies were incomplete and needed extensions, 
adaptation, and refinement as the technology hit the issues of the real 
world.   So in thinking about some alternate reality, one has to also 
think about how the technology might have evolved over the same time period.

To do that, IMHO the issue is not the technology per se.  Today's 
worldwide communications system could have been built using either OSI 
or TCP approaches, beginning with those 80s-era prototypes and evolving 
them into something quite different today.

So how might that have happened if TCP had just disappeared one day?

I think one of the important drivers of such evolution was the "culture" 
of the two worlds.   The TCP world was somewhat chaotic, with lots of 
ideas flying around and eventually congealing into "rough consensus and 
running code".   The managers and administrators at places like ARPA, 
NSF, et al fostered such an environment, explicitly labelling the work 
as "an Experiment", and encouraging new ideas that were unproven, 
unanalyzed, but might work.

In contrast, the OSI world was much more orchestrated, formal, and 
tightly managed.  I recall one instance in some Internet meeting where a 
discussion focussed on what the default value of some timing parameter 
should be.  Someone shouted out "How about 3 seconds?", and a consensus 
quickly formed that 3 seconds would probably be OK and we could always 
change it later.  In OSI land, a committee would likely have been 
formed, meetings held, and perhaps months or years later some 
recommendation would emerge, yet to be proven in actual deployed 
implementations.

I wasn't involved in much OSI work, but I do recall one meeting, 
somewhere in Europe about some technical topic that I can't recall.   
Progress was actually happening, as the proponents of various choices 
swayed each other toward a consensus, which seemed (to me) to be 
imminent.  Then the Chair interrupted, noting that it was now 3PM and 
therefore time to go on to the agenda item scheduled for that time.  So 
the productive discussion was halted, progress stopped, and the question 
was never resolved, as the meeting focus shifted to the new topic as 
scheduled.

In the TCP community, we would have continued that first discussion 
until consensus was reached, possibly adjourning to some restaurant in 
the interim, and watching the size of the group diminish until only the 
people who really cared about the result remained to hash out a 
solution.   Meetings in the TCP world also often had agendas, but I 
don't recall that we ever finished one as it was scheduled.

Other people have mentioned other aspects of the "culture" differences - 
e.g., the OSI tendency to focus on business models, competitive 
advantages, et al.   IMHO, all of these "cultural differences" would 
have had a significant effect on how OSI might have evolved into an 
alternate reality today.

One example might be videoconferencing.   We probably all have 
experienced today's ubiquitous videoconferencing over the Internet.   
How might that have evolved in an OSI world and how would we do 
videoconferencing today?

In the early 1990s, I was working in Silicon Valley and my company HQ 
used videoconferencing to interact with customers.   We had our own 
corporate "intranet" and were also connected to "The Internet", but 
neither of those had videoconferencing mechanisms available at the 
time.   There were experimental systems such as MBone, but those weren't 
usable for communications with customers outside of the research 
world.   But you could buy videoconferencing equipment and services from 
PTTs.

I don't recall the exact technical details, but in the 90s IIRC 
videoconferencing required use of two ISDN lines.  Those weren't normal 
everyday phone lines, so they were special orders to install the 
appropriate wires, modems,  monitors, cameras, et al into a room made 
for the purpose.   Using the ISDN lines was charged by the minute, but 
all the other costs dominated.   So we had only one special room set up 
for videoconferencing.   It worked amazingly well, providing clear and 
responsive video over just two 64 (or was it 56?) kilobits/sec 
circuits.  But of course few of the thousands of employees at our site 
ever used it, and not many customers had similar setups either.

What would such a system look like today if TCP had disappeared as it 
was expected to do and replaced by OSI?

Personally, I don't have a clue, but I can imagine lots of 
possibilities.  Given the slow pace of OSI's culture, we might be still 
using those ISDN lines.   Or perhaps the "TCP Culture" would have 
infiltrated the OSI community and produced 1000s of OSI RFCs despite the 
Chairs' attempts to exert control?  It might have happened either way.

But I think we can see some clues even today about how the OSI culture 
remains.   Just a week or so ago, I read a bit about the emerging "5G" 
cellular technology and in particular about its use of "Network Slices" 
to segregate traffic based on whether it is video, audio, interactive, 
bulk, etc.  One of the possibilities of using such a scheme is to apply 
different costs to each category.   So a video call might cost more per 
minute than an audio one.   All bits are not created equal.   Some are 
more lucrative than others.

In contrast, TCP-thinkers believe that all data is just bits.  It also 
seems that current thinking is that bandwidth is ubiquitous, free, and 
plentiful - although there are some anomalies like "data caps" that 
belie that thinking.

That wasn't always true.  Back in the early Internet, as TCPV4 was being 
defined, a "Type Of Service" (TOS) field was placed in each datagram 
header, reflecting a belief that there would be different kinds of bits 
that required different treatment as they were carried through the 
Internet.  Some bits, like those containing the next frame of video, 
aren't useful if they arrive too late to be used in creating the frame 
that was just displayed.

Over the decades however, the importance of TOS seems to have waned in 
the culture of today's Internet, with little (if any) attention now paid 
to TOS.  Functions used by mechanisms such as MBone haven't spread.  We 
have new terms such as "bufferbloat" to explain how today's Internet 
behaves.   All bits are created equal.  None are more important than 
others or need different handling.

So, if OSI had "won" and the technology had evolved to include features 
such as "Network Slices" what would videoconferencing today look like?

Perhaps it would still be expensive, requiring special rooms, equipment 
and ISP services, and as a consequence very limited in use - quite 
different from today's world where anyone, individual or huge 
corporation, can interact by video with their customers, families, 
friends, and colleagues.

Such interactions have become critical for many users of today's 
Internet.  But we have all gotten used to, and accept, the occasional 
glitches - the visual and audible dropouts, dropped connections, 
"buffering!" interruptions, and such characteristics of today's service.

In an OSI-based world, perhaps video uses would be more expensive and 
less pervasive.   But perhaps the video quality would have been 99.99% 
perfect.

Jack Haverty



-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20240315/0f1515fb/attachment-0001.asc>