[ih] History from 1960s to 2025 (ARPANET to TCP)

[Dave Crocker]

On 1/3/2026 1:20 PM, Craig Partridge via Internet-history wrote:
> I'm not going to try to answer the larger architectural question of why did
> things move into the host.   I wasn't there when the architectural
> decisions were made.


Many years ago I recall hearing Kleinrock comment about a pendulum swing 
of tradeoffs. I think it was about design choices when memory was 
expensive vs. cheap.  Or CPUs that were slow but got fast.  Or maybe it 
was about the tradeoffs caused by the combinatorics of these.  And 
communications speed vs. cost.  Or just speed.

In the mid-1980s, I worked at Ungermann-Bass, which is referred to as a 
LAN company but which marketed itself as a wire-replacement company, for 
connecting 'terminals' to 'mainframes'.  But it used LAN (and some WAN) 
tech to do this.

For connecting PCs, it did have a dumb Ethernet card, which I recall as 
being better than the very popular 3Com card.  Maybe it was, but they 
didn't (choose to / know how to) market it broadly.

But the workhorse it sold for PCs was a 'smart' card that had an Intel 
80186 on it, where they put the main networking data-handling code. so 
the user's PC wouldn't be burdened with that activity.  Because the 
dominate PC hardware was an Intel 8086, which was severely 
resource-limited.  The result was to make the PC's operational cost of 
using the Internet essentially free, rather than crippling it.

In 1970, 'mainframe's' needed to offload high-interrupt input/output 
tasks. Also in 1980 and maybe 1990.

By 2000, the issue isn't offloading, but making 'infrastructure' 
activity be independent of 'host' activity.

It's just a question of tradeoffs.


d/

-- 
Dave Crocker

Brandenburg InternetWorking
bbiw.net
bluesky: @dcrocker.bsky.social
mast: @dcrocker at mastodon.social