Joe Greco wrote on 31/03/2020 15:55:
There's a strange disconnect here. The concept behind Usenet is to have
a distributed messaging platform. It isn't clear how this would work
without ... well, distribution. The choice is between flood fill and
perhaps something a little smarter, for which options were proposed and
designed and never really caught on.
Without the distribution mechanism (flooding), you don't have Usenet,
you have something else entirely.
Exactly. And there's no disconnect: usenet doesn't scale because each
object is copied to all core nodes rather than referenced, or
copied-as-needed, or other. This design of distributed messaging
platform will eventually break as it grows. It's ok to acknowledge this
explicitly: message buses are useful, but they have their limits.
Kinda like how there's a problem with the technology of the Internet
because if I wanna be a massive network or a tier 1 or whatever, I
gotta have a massive investment in routers and 100G circuits and all
that? Why can't we just build an Internet out of 10 megabit ethernet
and T1's? Isn't this just another example of your "problem with the
technology at the design level?"
No, not even slightly. Is an NSP expected to carry all traffic for the
entire DFZ? Because that's your proposed analogue here.
Usage grows. I used to run Usenet on a 24MB Sun 3/60 with a pile of
disks and Telebits. Now I'm blowing gigabits through massive machines.
This isn't a poorly designed technology. It's scaled well past what
anyone would have expected.
yeah, I don't miss those days. I ran news on a decsystem 5100 with a
couple of megs of RAM and a single disk. My desktop was a sun 3/60. Not
pretty, but at least it could fit 4 xterms on-screen. In that sense, it
was almost as functional as my ragingly fast desktop is these days.
Nick