22 Feb
2026
22 Feb
'26
11:37 a.m.
Paul Koning <paulkoning(a)comcast.net> wrote:
On the other hand, the even more obscure DECnet/8 was found a couple
of years ago. Its main limitation ....
Besides being Phase-I, DECnet/8 is not a "complete" implementation. The
only application provided is a console to console talk program that can only
talk to other PDP-8 nodes. There's no FAL, CTERM/RTERM, NFT, NCP, or any of
the other things you might expect. I think DEC expected customers to write
their own RTS/8 and DECnet-8 applications.
Bob
22 Feb
22 Feb
12:09 p.m.
New subject: DECnet-RT
On Feb 22, 2026, at 2:37 PM, Robert Armstrong via
cctalk <cctalk(a)classiccmp.org> wrote:
True, but most of those were not defined yet. Certainly no CTERM, and NCP didn't
reallly exist until Phase III (though TOPS-20 Phase II does have an NCP, with syntax
different from the later one matching no specification I have ever seen).
The object number registry suggests there was an early DAP, predating the "5.6"
version implemented in DECnet/E, but I have never seen a spec for that either.
TLK/LSN is not quite compatible with the version in other DECnet implementations but
close, and tweaking an implementation to support both flavors of the protocol isn't
hard.
paul
Paul Koning <paulkoning(a)comcast.net> wrote:
On the other hand, the even more obscure DECnet/8 was found a couple
of years ago. Its main limitation ....
Besides being Phase-I, DECnet/8 is not a "complete" implementation. The
only application provided is a console to console talk program that can only
talk to other PDP-8 nodes. There's no FAL, CTERM/RTERM, NFT, NCP, or any of
the other things you might expect. I think DEC expected customers to write
their own RTS/8 and DECnet-8 applications.
Bob 
11:05 p.m.
New subject: DECnet-RT
Paul Koning wrote:
NCP didn't reallly exist until Phase III (though
TOPS-20 Phase II does
have an NCP, with syntax different from the later one matching no
specification I have ever seen).
What does NCP mean in this context? I thought DECnet NCP meant Network
Control Program, same as in an ARPANET, CHAOSnet, or SNA context.
I.e. a software component now known as "network stack".
23 Feb
23 Feb
12:16 a.m.
New subject: DECnet-RT
NCP is the user interface for DECnet configuration, not the network stack.
It is the "Network Control Program", the program to control DECnet
networking.
For example:
NCP> SHOW KNOWN NODES
will show the node known to the node you execute this command on.
Ulli
Am Mo., 23. Feb. 2026 um 09:09 Uhr schrieb Lars Brinkhoff via cctalk <
cctalk(a)classiccmp.org>gt;:
Paul Koning wrote:
NCP didn't reallly exist until Phase III
(though TOPS-20 Phase II does
have an NCP, with syntax different from the later one matching no
specification I have ever seen).
What does NCP mean in this context? I thought DECnet NCP meant Network
Control Program, same as in an ARPANET, CHAOSnet, or SNA context.
I.e. a software component now known as "network stack".
6:18 a.m.
New subject: DECnet-RT
Correct.
The layer and component names in DECnet are different from those in other protocols, OSI
as well as TCP/IP. Phase IV tried to move somewhat towards OSI terminology and of course
Phase V completed that move, but Phase III and before are different. So we have NSP which
OSI calls "transport" and ARPAnet calls TCP (and before that NCP, I think). And
Phase III speaks of the "transport" layer which OSI calls the network layer and
ARPAnet calls IP.
Another somewhat confusing thing is that in some cases DECnet has conventional names for
the client, the server, and the protocol, all separate. So we see NFS/FAL/DAP,
NCP/NML/NICE, TLK/LSN/<something>.
NCP as a standardized element, with its protocol NICE, arrived in Phase III. I was
somewhat surprised to find a predecessor of both the program and the protocol in Phase II
for TOPS-20, since as far as I remember there isn't anything like it in the Phase II
version of DECnet/E, nor have I ever seen a protocol spec for the Phase II version of
NICE. Reverse engineering it from the sources explains the odd message numbering in
(phase III) NICE, though: the numbering starts at 15 because the numbers below that were
used (mostly for similar functions but with quite different encoding) in the Phase II
version of the protocol.
paul
On Feb 23, 2026, at 3:16 AM, Hans-Ulrich Hölscher via
cctalk <cctalk(a)classiccmp.org> wrote:
NCP is the user interface for DECnet configuration, not the network stack.
It is the "Network Control Program", the program to control DECnet
networking.
For example:
NCP> SHOW KNOWN NODES
will show the node known to the node you execute this command on.
Ulli
Am Mo., 23. Feb. 2026 um 09:09 Uhr schrieb Lars Brinkhoff via cctalk <
cctalk(a)classiccmp.org>gt;:
> Paul Koning wrote:
>> NCP didn't reallly exist until Phase III (though TOPS-20 Phase II does
>> have an NCP, with syntax different from the later one matching no
>> specification I have ever seen).
>
> What does NCP mean in this context? I thought DECnet NCP meant Network
> Control Program, same as in an ARPANET, CHAOSnet, or SNA context.
> I.e. a software component now known as "network stack".
>
6:26 a.m.
New subject: DECnet-RT
On Feb 23, 2026, at 9:18 AM, Paul Koning via cctalk
<cctalk(a)classiccmp.org> wrote:
Correct.
The layer and component names in DECnet are different from those in other protocols, OSI
as well as TCP/IP. Phase IV tried to move somewhat towards OSI terminology and of course
Phase V completed that move, but Phase III and before are different. So we have NSP which
OSI calls "transport" and ARPAnet calls TCP (and before that NCP, I think). And
Phase III speaks of the "transport" layer which OSI calls the network layer and
ARPAnet calls IP.
Another somewhat confusing thing is that in some cases DECnet has conventional names for
the client, the server, and the protocol, all separate. So we see NFS/FAL/DAP,
NCP/NML/NICE, TLK/LSN/<something>.
NCP as a standardized element, with its protocol NICE, arrived in Phase III. I was
somewhat surprised to find a predecessor of both the program and the protocol in Phase II
for TOPS-20, since as far as I remember there isn't anything like it in the Phase II
version of DECnet/E, nor have I ever seen a protocol spec for the Phase II version of
NICE. Reverse engineering it from the sources explains the odd message numbering in
(phase III) NICE, though: the numbering starts at 15 because the numbers below that were
used (mostly for similar functions but with quite different encoding) in the Phase II
version of the protocol.
More: in Phase IV we recognized the naming conflict between existing DECnet layers and
OSI. Rather than go directly to OSI names, that was judged to be confusing because the
two most important layers have roughly the same names but reversed: transport = network
layer, and NSP (network services protocol) = transport layer. So instead Phase IV was
given new names for those layers that are descriptive and were not in use for any other
purpose at the name: "routing" and ECL ("end communication layer").
paul
3 Mar
3 Mar
7:08 p.m.
New subject: DECnet-RT
Paul Koning wrote:
NCP as a standardized element, with its protocol NICE,
arrived in Phase III. I was somewhat surprised to find a predecessor of both the program
and the protocol in Phase II for TOPS-20
I'm 99.44% sure that in Phase III days both PDP-10 operating systems
were still non-routing Phase II nodes with only one link to a
dedicated front end PDP-11 (via the DTE hardware that could handle
upto four). Before Phase III everything had to be manually routed,
A::B::C -- I forget the name of the passthru service that facilitated
that.
My memory is that the software running on the '11 was called MCB.
Ah, looks like I've posted similarly in the past on the SimH list:
https://groups.io/g/simh/topic/decnet/105849791
5 Mar
5 Mar
7:32 a.m.
New subject: DECnet-RT
On Mar 3, 2026, at 9:36 PM, Phil Budne via cctalk
<cctalk(a)classiccmp.org> wrote:
Paul Koning wrote:
The front end thing is called "intercept" which implements a rudimentary form of
routing documented in an appendix of the Phase II spec. PyDECnet implements it (in a
branch at least; I forgot if it's released yet).
But that doesn't use the A::B::C stuff you mentioned. That's a different
mechanism handled in the application (in some of them) using a DEC internal object called
PMR (poor man's routing) or PSTHRU. That actually was carried over all the way to
Phase IV within DEC in order to access "hidden areas".
Some applications would get equivalent functionality a different way. For example, it
would work in DAP if the intermediate node was VMS, because VMS (RMS32) has transparent
network access.
paul
NCP as a standardized element, with its protocol
NICE, arrived in Phase III. I was somewhat surprised to find a predecessor of both the
program and the protocol in Phase II for TOPS-20
I'm 99.44% sure that in Phase III days both PDP-10 operating systems
were still non-routing Phase II nodes with only one link to a
dedicated front end PDP-11 (via the DTE hardware that could handle
upto four). Before Phase III everything had to be manually routed,
A::B::C -- I forget the name of the passthru service that facilitated
that.
14
days inactive
25
days old
7 comments
5 participants
participants (5)
-
Hans-Ulrich Hölscher -
Lars Brinkhoff -
Paul Koning -
Phil Budne -
Robert Armstrong