Re: IPv6 Anycast has been killed by LINUX patch in 2016 - who cares?

[Töma Gavrichenkov <[e-mail]>]

Peace,

On Sun, Aug 8, 2021 at 1:09 AM Robert Raszuk <[e-mail]> wrote:
> But using connection less forwarding plane and transporting connection oriented protocols (like TCP) only works between solid anchors (read endpoints).

Tell you what.
Some 12 years ago I was working in HP, and I had to do some weird
things with OpenVMS.

OpenVMS (formerly VMS, and VAX/VMS before that) is/was a true real
time system.  Lots of life-controlling systems were built on top of
OpenVMS these days.  I recall the ISS is still flying because it's
powered by VAX/VMS but I might really be wrong about it.

Anyhow, the TCP protocol wasn't considered reliable in VAX/VMS because
it didn't guarantee reliable data delivery.  By "reliable", the VMS
architects meant the reliable delivery over the reliable maximum
timeframe.

This reliable delivery required actual network and transport network
architecture to be worked on.  Copper line maps from every endpoint to
every endpoint were a part of every project, with redundancy ensured,
and network continuity protocols were signed off by the customer, with
multi-country regulation entities also having their signoffs,
guarantees, and insurances.

And it was expensive as hell, but it worked just the way you want it
to work: stable, reliable, and constant over time.

The Internet isn't that expensive because it's built on compromises.
TCP is one of these compromises.  It doesn't guarantee reliable
delivery in terms of VAX/VMS — and neither do SCTP nor QUIC —  but all
of these are cheap, and the rest is up to the application, and it
should be the application that could handle that.

Yet, no one should never expect a perfect performance from a
compromise.  It just was never paid for.

--
Töma