I don't have time to dig up the details. But TOPS-10/20 definitely supported the CI. It turned out not to be used much since IIRC despite the high wire speed, thruput was better on the NI. But the rationale was something like if you had CI connectivity for storage, and an NI (ethernet) issue, you had to have a means of managing the other cluster nodes. So DECnet of CI ensured you could set host, do task-task; whatever.
For some evidence, look at ntman.mac: TABSR1: TXNN T1,<.NTLIN&.NTCKT> ;Is it a LINE or a CIRCUIT? JRST TABSR2 ;No, don't bother checking Line type LOAD T2,NXLTY,+NMXVAR ;Get the Line type MOVE T2,[NTD.D ;Bit indicating DTE NTD.K ; KDP NTD.P ; DDP NTD.C ; CI NTD.N ; NI NTD.R]-1(T2) ; DMR TDNN T2,NT.DEV(NT) ;Does this Parameter apply to this Line Type? NMXN23: STOR T1,LIDEV,+P2 ;STORE IN CIRCUIT ID WE ARE BUILDING CALL NMXN28 ;GET A NUMBER FROM THE STRING STOR T1,LIKON,+P2 ;STORE AS KONTROLLER NUMBER IN CIRCUIT ID CALL NMXN28 ;GET THE NEXT NUMBER FROM THE STRING STOR T1,LIUNI,+P2 ;SAVE UNIT NUMBER CALL NMXN28 ;GET NEXT NUMBER (IF ANY) STOR T1,LIDRP,+P2 ;SAVE AS DROP NUMBER (PORT ON CI) MOVE T1,P2 ;GET THE CIRCUIT ID WE JUST BUILT RETSKP ;RETURN SUCCESS. ;WE KNOW THAT ALL DEVICE NAMES ARE AT MOST 4 BYTES LONG, AND THUS FIT IN A WORD HRLI T2,(POINT 7,) ;WE STORE THE DATA IN 7 BIT BYTES INTERNALLY HRRI T2,KONNAM(T1) ;GET POINTER TO DEVICE NAME NMXC20: ILDB T1,T2 ;GET A BYTE FROM THE DEVICE NAME JUMPE T1,NMXC21 ;END OF STRING, FALL THROUGH IDPB T1,P2 ;SAVE IT IN OUR DESTINATION DATA STRING AOJA T6,NMXC20 ;INCREMENT NUMBER OF BYTES DEPOSITED NMXC21: MOVEI T1,"-" ;SEPERATOR IDPB T1,P2 ;STORE IT AOJ T6, ;COUNT UP CHARACTER LOAD T1,LIKON,+P1 ;GET THE KONTROLLER NUMBER PUSHJ P,NMXC28 ;OUTPUT NUMBER IN DECIMAL MOVEI T1,"-" IDPB T1,P2 ;STORE THE SEPARATOR AOJ T6, ;INCREMENT NUMBER OF BYTES LOAD T1,LIUNI,+P1 ;GET THE UNIT NUMBER PUSHJ P,NMXC28 ;OUTPUT THE NUMBER IN DECIMAL JN LILXC,+P1,NMXC22 ;Jump if we are not dealing with a CIRCUIT LOAD T1,LIDEV,+P1 ;GET THE DEVICE TYPE CAXE T1,LD.CIP ;IS IT A CI? JRST NMXC22 ;NO, WE ARE DONE MOVEI T1,"." ;OUTPUT SEPERATOR IDPB T1,P2 ;STORE IT AOJ T6, ;AND INCREMENT COUNT LOAD T1,LIDRP,+P1 ;GET THE DROP NUMBER (PORT ON THE CI) PUSHJ P,NMXC28 ;AND OUTPUT IT IN DECIMAL On 09-Mar-17 13:00, Paul Koning wrote: >> On Mar 9, 2017, at 12:27 PM, Timothe Litt <l...@ieee.org> wrote: >> >> .. >> With respect to an earlier question about CI emulation: The best approach >> may be to emulate it as multicast ethernet. Setup a couple of groups for >> each star coupler. Each coupler supports up to 32 nodes. For VMS, half o >> them hosts, half storage. CI is a 50Mb/s (might be 70, but I remember 50 >> from the initial doc) CSMA/CD bus. > 70 Mb/s is the number I always saw. > >> The cabling is redundant (A & B cables), so assign one MC group to each per >> star. (Yes, there are DECnet drivers for on multiple OSs. I think it was >> considered a multipoint medium with addresses related to (CI) node numbers >> rather than an ethernet (broadcast) - but it's been a while.) > I don't think DECnet ever supported CI. It certainly doesn't show up in the > DECnet architecture specs anywhere. > > paul >
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ Simh mailing list Simh@trailing-edge.com http://mailman.trailing-edge.com/mailman/listinfo/simh