28 Oct
2021
28 Oct
'21
10:33 p.m.
On Oct 28, 2021, at 3:35 PM, John Forecast <john at
forecast.name> wrote:
Interesting. I wonder what it does to recover from packet loss due to CRC error.
paul
On Oct 28, 2021, at 3:10 PM, Paul Koning
<paulkoning at comcast.net> wrote:
I just checked the DECnet-RSX 4.2 SPD and it explicitly says that the PCL11-B has its own
line protocol and does not use DDCMP.
John. On Oct 28, 2021, at 2:55 PM, John Forecast
<john at forecast.name> wrote:
The PCL-11B was a CSS product which allowed up to 16 Unibus PDP-11 to be connected
together over a TDM bus running at 1 megabyte/sec. I?m pretty sure it didn?t use DDCMP.
I?m not sure about Phase II but for Phase III each node was set up as a multi-point
master (PCL-0.0) and the remote nodes were multi-point tributaries (PCL-0.1 to PCL-0.15)
so it would appear as a fully connected mesh. At the end of Phase III, Hardware
Engineering built a small number of prototype Unibus ethernet boards (maybe 10 total) so
they could get some experience with ethernet (this was a programmed I/O device with 1
transmit and 1 receive buffer, sort of like some of the early 3Com devices). I used a
similar technique to the PCL to get RSX-DECnet Phase III running (slowly) over ethernet.
It looks like the PCL does CRC checking, and it's a multipoint device. So that's
a bit like the lower half of a DMP-11.
The DDCMP spec explicitly mentions support for parallel links. And it makes sense for
PCL -- given a CRC checker in the hardware, the other part you want is timeout and
retransmission. DDCMP provides that. The circuit naming you described is the same as
what DECnet Phase III introduced for supporting the DMP-11.
paul