[ih] GOSIP & compliance

[Clem Cole]

On Sat, Mar 19, 2022 at 12:56 PM Craig Partridge <craig at tereschau.net>
wrote:

>
>
> On Sat, Mar 19, 2022 at 9:00 AM Clem Cole via Internet-history <
> internet-history at elists.isoc.org> wrote:
>
>>
>> *IP vs. OSI    -- "**Simple Economics always beats Sophisticated Design"*
>>
>>
> I would actually says "works beats does not work."  Not because OSI
> couldn't work.  I think, for the most part, the implementation efforts of
> the time showed that, with some set of bugfixes and adjustments to
> standards, OSI could work.  But... TCP/IP was already working.
>
Exactly ... economics is the high order bit.  Doesn't matter if its
legislated or not.   If it works and gets the job done, is providing value,
its pretty hard to displace it [hey IPv6 has yet to displace IPv4 in
practice for the same reason -- it works and is economical].

Cristinsen's book explains it.  To successfully disrupt, you have to find a
new  (and rapidly growing) user base that values the new technology AND is
willing to accept its downsides at the beginning.  But Metcalfe notes
that's really hard in communications networks, because the value of the
network is less determined by the technology, but by the number of users
that are part of the community.

>
> As best I can tell (I didn't join the scene until 1983), 1990 OSI
> implementations were not as mature as 1981 TCP/IP implementations.
>
I agree.



>   And you would not have wanted to run a 1981 TCP/IP network -- indeed, a
> certain share of ARPANET folks were none too happy when forced to run
> TCP/IP in 1983.  Several years of operational experience made a huge
> difference in terms of operational stability.
>
Metcalfe's law.   There was not a new user base and the old user base
valued what it had.  It >>just worked<< and the network was growing at an
incredible rate making it even more valuable if you joined it.

Note, it is not as if the OSI advocates did not know this.  In fact, when
> queried, they'd say they needed to import the wisdom of TCP/IP operations
> into OSI.  But... they never did (and there's probably a good case study
> there about why).
>
I'm not so sure if they had just imported it, it would have been enough to
displace it.   Politics and business interests aside, you had a lot of
smart techies on all sides.  But this was an economic issue to the end user
[and network operator].  The OSI folks either needed a whole new set of
customers to create a new network that the IP folks were going to want to
flock too (unlikely IMO), or they needed to find a way to use the Microsoft
'Embrace and Extend' idea -- join the mainstream and then make something of
value that was only possible in their world.  Which (as MSFT discovered)
basically went against the grain of the way IP was built/maturing.