[truantbuick]

Isn't this article missing the point of the internet?

It talks about an ambitious (but ultimately aborted) top-down project to build thousands of interlinking mainframes.

But building large networks wasn't particularly novel. It was the idea that you could build a logical layer that potentially linked any network.

The key to the early internet was you didn't necessarily need to build anything physical. You could link up several existing networks in any which way you want, despite involving disparate organizations, systems, and infrastructure.

[aglavine]

Also, the key to the early internet was the concept of packet-switching.

[boudin]

CYCLADES was really influencial on that (https://en.m.wikipedia.org/wiki/CYCLADES)

[emiliobumachar]

Could you please elaborate?

How would one link up two existing networks with different systems and infrastructure?

What I can think of is either installing a couple translating routers, which do speak TCP/IP as well as their network's internal protocol, or making them, by repurposing existing machines with software-only modifications. Is that what you had in mind as not really counting as installing something physical, or did I miss something?

[thodin]

Modern "Internet" is about about making everything tcp/ip based. Interconnection between networks with different protocols was a real thing long before tcp/ip conquered our planet:

"An amusing side note on the VNIIPAS connection: while the author of this paper was in Havana, he connected to a VAXNMS system at his home via the following path: PAD program on Unix microcomputer at CENIAI in Havana goes over X.25 board local to that system; X.25 line from Havana to Moscow, via satellite; VNIIPAS X.25 data switch receives call, routes to international Sprint network via Western Europe; Sprint carries call through some number of cities and links to Reston, Virginia where it conveys call to Columbus, Ohio, to CompuServe's X.25 gateway; CompuServe carries call from Columbus, Ohio to Tucson, Arizona, where it gets translated from X.25 formats to internal DECnet format and passes over the University of Arizona DEC net network, through Ethernet, fiber optic, 56 Kbps synch and asynch 19.2 Kbps TCP/IP lines to author's home over another Ethernet from gateway to workstation, returning with the prompt "Username:" The miraculous thing about this call is that it was done with a single X.121 address at the Havana end." - this is how it was in 80s.

[Eyas]

Right. The key innovation of the internet was the invention of TCP/IP for the purpose of inter-network communication. See https://en.wikipedia.org/wiki/Internet_protocol_suite#Early_...

[ethbro]

I'd say the kernel of that was the idea of encapsulation and separating functions by layer (really the same idea, from packet and flow perspectives, respectively).

[LoSboccacc]

I'd say the core is that anyone could link up and start talking with other parts of the network; the arp/ip technology was the stepping stone of course, but a key differentiator from previous machine networks[1] is that you can just hook up to the internet, it was truly decentralized from its ideation.

[1] ie. stock tickers, terminals of time share mainframes, or see plan 55-A for a wider interconnected switched network this russian idea