On 2006-08-01 00:44:49 +0100, "Doctor Drivel" said:


"Andy Hall" wrote in message ...
On 2006-07-31 23:16:12 +0100, "Doctor Drivel" said:

"Andy Hall" wrote in message ...

Matt...no...... Conectionless was one of the key points.

Once they realised how dreadfully slow connection oriented was and that
it was unnecessary anyway.

Matt, no. It was in the original concept.

Wrong.



That would certainly have killed it if the committees hadn't.

The committees had scum like IBM on them, which they should not have had.

There were all kinds of people on them. The problem was that through
having to use lengthy and bureaucratic procedures, progress was
incredibly slow.

Nope. The IBMers slowed everything down nit-picking a small point in a
layer for ever more. They deliberately were dragging their feet.

Nit picking is the inevitable outcome of committees, government
involvement and attempts at international standardisation. The
procedures are inevitably lengthy. It is not reasonable to lay that at
the door of IBM or any other individual participant.



The market moved on and left them behind.

The market was way behind. Only the Internet and www took off like a
rocket, mainly the www. That sealed it as there was too much of the
TCP/IP crap around.

Uncle Sam was going OSI.

Not for very long.

After the www skyrocketed, Sam dropped it.

Not quite. The main reason was that the OSI protocol suite was going
to be "two years away" since about 1986. The U.S. government DoD
attempted to mandate it in the form of GOSIP for all environments from
1990 onwards. That never happened and so the first bodge was to come
up with something that would allow OSI to be run over TCP/IP.
Realistically that was never going to work in any useful way and so
attempts were made to incorporate TCP/IP features into OSI - TP4 is one
example of that (note that it can run on a *connectionless* network
layer. Finally, NIST suggested that the DoD drop the "OSI only"
requirement in 1994.

Essentially, this was a political fudge that saved red faces among
those that had embarked on the OSI bandwagon, allowing them to change
position.

WWW substantial growth was from 1993 onwards, and may have been one
contributor to the demise of OSI in US government use, but it is clear
that the main reason was the continuing delays and lack of
deliverability of anything from the OSI committees.




It was used by BMW extensively, British government departments used it
too, along with European and US..

Quite a number of large companies used X.25 because that was a standard
telco offering and could be used internally as well.

It worked, was seamless and was fast enough for the time. The all knew
it had to go, and frame relay (X.24 without the checking to make it
faster), etc were implemented too. Many organisation are still on X.25
and replaced with faster hardware. They have no problems at all on
faster speeds and have no desires to change over.

Rubbish. There is very little X.25 left in corporate networks any
longer. There is still quite a lot of frame relay, but that is
disappearing rapidly as telcos migrate customers to alternative
technologies.




There was an initial assumption that it would migrate, in modified form
to more substantial networks. This never went anywhere because
progress was too slow and the market passed it all by.

The rapid spread of the Internet and the w.w.w., which had not adopted
OSI as it was still being implemeted in various government
departments and had not quite reached the rest, killed OSI.
Nothing else. It was too late to turn back the TCP/IP protocol. If the
www had been two years later it probably would have had an OSI protocol
stack.

No it wouldn't.

It would have.

No it wouldn't. IP based networks were well established before Mr
Berners-Lee came along.

You are slow. They were to be replaced by OSI in major organisations
and makers would push OSI too, then private users would adopt OSI as
they went along, but www/Internet used TCP/IP.

This was all completely theoretical because of lack of performance and
deliverability.


Companies like REtix had off the shelf OSI stacks for ethernet, token
ring, token bus, for UNIX boxes and PCs runing Windows too (well DOS
then was doing the work).

All of which is irrelevant because it never went anywhere.



OSI wasn't implemented fast enough because the Internet wasn't
regarded as that important at the time. The www made it important.
Before that it was for nerds and fellas with beards and mussies.

OSI wasn't implemented fast enough because of the standardisation
approach that was taken. Realistically, it was doomed to failure
from the outset. I can remember making that prediction in about 1988
or 89 when there were a few government tenders around looking for GOSIP
implementation. TCP/IP adoption was already well underway in the
commercial world from about 1986 onwards. One could argue that WWW
growth was a significant factor later on, but I can't think of anybody
seriously intending to implement OSI from about 1988 onwards. The
death knell had sounded long before 1993 for OSI in terms of an
alternative to TCP/IP.