[HECnet] GRE Peering

[Paul Koning]

> On May 21, 2019, at 5:12 AM, Keith Halewood <Keith.Halewood at pitbulluk.org> wrote:
> 
> Hi Johnny,
> 
> That's fine by me.
> In the cases where the other side has a fixed IP address, I would prefer to make the TCP connection. I can keep my firewall rules simple and in failover situations my range of IP addresses can change. Where the other side has a dynamic IP address, they'll obviously have to initiate the TCP connection but I can at least limit acceptable incoming IP addresses down to a range.
> 
> Does anybody have any experience of running DECnet circuits over Multinet's (default) UDP configurations?

Yes, in tests at least.  It's not a good solution.  The reason is that Multinet is, from the DECnet point of view, a "point to point datalink" and the DECnet requirements for such a link are NOT satisfied by Multinet over UDP.  In other words, to put it bluntly, Multinet over UDP is a DECnet protocol violation.

With a bit of luck and patience the connection typically does come up and can be used.  But if you want something that's reliable, better choices are (a) Multinet over TCP, (b) GRE, (c) DDCMP (as implemented by SIMH).  (b) and (c) at least clearly obey the architectural requirements, and (a) either does or is close enough to be reasonable.

	paul