[ih] History from 1960s to 2025

[Craig Partridge]

Hi Jack:

Much of the short summary seems roughly right to me.  I'd suggest, however,
that the following misses a lot:


> - 1980s: US government embraces COTS (Commercial Off The Shelf) policy,
> which encourages the development of commercial products for use in the
> TCP environment;  corporate representatives from tech companies begin to
> participate in Internet technology development and standardization
> efforts (IETF); DoD limits funding of custom systems and research in
> favor of using commercial products
>
> - 1990s: Commercial users, and the public, get tired of waiting for the
> internet wars to end, notice that TCP technology is available, can be
> observed to work, and can solve their immediate IT problems; the TCP
> Internet grows rapidly in the general public worldwide; corporations
> deploy private "intranets" using TCP products; all competing internet
> architectures fade into oblivion
>

Specifically, between late 1987 and 1993, so straddling the decade boundary
there were an intense 5ish years in which TCP/IP crashed through a bunch of
scaling/deployment/commercialization issues.  (Matured from a network that
could support a few tens of thousands of users to one that could support
billions).   That period saw the first working congestion control (VJCC -
the one everyone now gripes suffers bufferbloat, but it effectively fixed
congestion control for two decades), BGP replacing EGP (which failed early
and messily), IS-IS and OSPF (the intra-domain routing protocols such as
RIP did poorly), DHCP (RARP didn't cut it...), IPsec (ability to have
secure tunnels), as well as things like MIME (fixed email to work with
international character sets -- and, bonus, gave us graphics and proper
attachments).  The community by 1993 was also well on its way to solve
TCP's challenges on gigabit speed links.

This flood of solutions meant that when commercial users and the public
"tired of the internet wars to end" (and frankly, I think most folks
weren't even aware protocol wars existed until they concluded a corporate
Netware infrastructure no longer met their needs), when they looked, there
was one solution that (a) had 90% of what they needed; (b) demonstrably
worked at massive scale; and (c) didn't lock them into a vendor (e.g.
DECNet was close, but...).  Specifically (a) and (b) were not true in 1987
of TCP/IP but very much were true by 1993.

I've argued elsewhere (a paper on Internet Governance published years ago
in IEEE Annals) that as the Internet dramatically matured in this short
period, the understanding of that maturation propagated unevenly through
the technical and policy worlds, in ways that caused stresses.


Craig

-- 
*****
Craig Partridge's email account for professional society activities and
mailing lists.