[HECnet] This is probably been asked already but....

On Jul 1, 2012, at 9:33 PM, Johnny Billquist wrote: On 2012-07-02 03:17, Gregg Levine wrote: On Sun, Jul 1, 2012 at 9:10 PM, Bob Armstrong <bob at jfcl.com> wrote: ... What I am curious about is how the PDP-11s talked to DEC Net based systems. And which ones of the R* operating systems could be confused into doing that, and in what ways. Well, considering the fact that DECnet was born on the PDP-11, the question should perhaps be rephrased as how non-PDP11 systems managed to talk on DECnet. :-) If the question is more on a practical note of how connections were done, I think the original was over serial lines, but I might be wrong. Multidrop as well as some high speed links were available in Phase III, but ethernet only became an option with DECnet phase IV. Not sure about the time frame for packet switched interfaces. They certainly existed in Phase IV, but I don't know if they were available already in phase III. Let's see. Some history, as best as I remember it. This is partially before my time, and also partially derived from on-line history files... DECnet Phase I was implemented on RSX. Not sure which flavor. I'd guess RSX11-M just because that's where the action was. DDCMP only, in software only as far as I recall. Probably on a DU-11. DECnet Phase II was the first multi-platform DECnet. I suspect it's also the first that had formal hardware-independent protocol specifications. Protocols were DDCMP for the data links (point to point only, sync or async), NCP, and DAP. Also around that time some other things appeared; I think MAIL-11 (email). I'm fairly sure that predates DECnet Phase III because I vaguely remember it using UUCP style paths (foo::bar::pkoning) in that time -- which later came back for PMR support. Supported operating systems included RSX, RSTS, VMS, TOPS-10, and I think TOPS-20. Terminal emulation appeared, but in an OS-specific fashion that was basically undocumented. Network management (NCP and NML) was now part of the standard architecture and fairly consistent among the various implementations. DECnet Phase III introduced routing, which made it useable for networks the size of the DEC internal network ("Engineering network" or E-net for short) by introducing routing: one area, 255 nodes max. Still no Ethernet, but this is when multipoint DDCMP was added. That was a bit of a dancing bear ("The amazing thing is not how well the bear dances, but that it dances at all") in my view, but it did work after a fashion. It may be that packet switching (X.25) appeared here -- I never knew much about that since it wasn't ever considered for RSTS and was essentially nonexistent in the USA as far as I could tell. Check the Phase III routing layer spec ("version 1.3") and/or the Phase III network management layer spec -- if either can be found it will certainly answer this question. DECnet phase IV adds areas, and increases the bottom level routing to 1023 nodes max by creating a way to break up the routing message into pieces (and send only what changed, as opposed to the data for the whole network every time as Phase III did). And, most importantly, it adds Ethernet. The Ethernet packet formats are strangely different in the routing layer; this is a leftover from an earlier proposal for "Phase IV" which was intended to introduce long addresses and link-state routing; that was rejected by the implementers as being too hard and "Phase III Extended" was whipped up instead by a slight tweak of Phase III to make it hierarchical. Parts of the original "Phase IV" eventually ended up in Phase V, though not the addressing; that was replaced by OSI addressing. But things like the "lollypop sequence numbering" in the Phase V routing protocol (and in IS-IS, which came directly from there) carried over from the earlier proposal. BTW, note that OSPF was derived from DECnet Phase V routing, though the OSPF creators never bothered to give any credit for that. Supported platforms: eventually nearly all of the Phase III platforms, though some (like RSTS) took quite a while to get there. BTW, high speed links -- by the standards of the time -- appeared around Phase II when the DMC-11 was introduced. It came in two flavors, one to support lines with modem control at up to 56 kb/s (actually, I suspect they would run somewhat faster, if you could find a modem to run faster) and the other for service within a building, running over coax at 1 Mb/s. By the standards of the late 1970s that was seriously fast, especially for PDP11 based minicomputers. It also was a strain for the microcontroller in the DMC-11, which is why that one speaks a somewhat buggy version of DDCMP. I may still have the details somewhere; it wasn't all that hard to make a standards-compliant DDCMP implementation work with a DMC-11 but it takes a few hacks here and there. paul