[HECnet] Intermittent Connection with PyDECnet?

[Johnny Billquist]

The parsing in RSX parse a whole bunch of things that cannot be set, or 
sometimes even exists in RSX. The thing is, the same parsing is also in 
the path of processing commands for remote nodes, where those things 
might make sense.

In fact, muhc processing under RSX is always done over NICE, even for 
the local host. So in fact, the parsing side do not know what actually 
is possible, or makes sense. In the NICE protocol, the tokens are 
defined, and NCP just creates the request for the NICE server to 
process. The RSX NICE server gives errors back to anyone who tries to 
change something the RSX DECnet don't allow. NCP does not.

   Johnny

On 2020-03-03 01:54, Thomas DeBellis wrote:
> Yes, this is /exactly/ what I was talking about below, "I sometimes get 
> the impression that it parses for more than it can actually do."  You 
> can put any kind of circuit id you want, but you get the exact same pick 
> list, even when the parameter makes no sense.  NMLT20 does the semantic 
> processing and decides whether the request passes muster.  So it looks 
> like I don't have it, but I'll have a look at the sources maybe later.
> 
> It's not a limitation of Tops-20 but rather that the OPR user mode 
> parser does not have the appropriate semantic knowledge.
> 
> You can get some really fancy parsing with COMND%.  I guess my favorite 
> was a program I wrote that built its parameter list on the fly based on 
> what you were able to set and what you had already set.  So, you could 
> type a "?" and be assured that what was listed was in fact settable.  As 
> you set each parameter, it got taken off the parse list, so you could 
> get real time feed back on what you had left to do.
> 
> Very nice and a bit overboard, but that's me, I guess.  I have never 
> seen COMND%'s equal anywhere on Unix or Windows or anywhere else except 
> for C-MM and C-KERMIT, both of which used Columbia C based COMND% package.
> 
>> ------------------------------------------------------------------------
>>
>>  On 3/2/20 7:36 PM, Robert Armstrong wrote:
>>
>> FWIW, you can’t/couldn’t do verification on an Ethernet circuit 
>> anyway.  It only works on point to point links.  Command completion 
>> may be smart enough to only show you the options that are applicable 
>> to your input so far, or maybe TOPS20 just doesn’t have it…
>>
>> Bob
>>
>> *From:*owner-hecnet at Update.UU.SE [mailto:owner-hecnet at Update.UU.SE] 
>> *On Behalf Of *Thomas DeBellis
>> *Sent:* Monday, March 2, 2020 3:55 PM
>> *To:* hecnet at Update.UU.SE
>> *Subject:* Re: [HECnet] Intermittent Connection with PyDECnet?
>>
>> I appear to have neither CIRCUIT VERIFICATIONnor NODE TRANSMIT 
>> PASSWORD, viz:
>>
>>     NCP>set cirCUIT ni-0 ? ALL
>>       or one of the following:
>>      ACTIVE         BABBLE      BLOCKING     CHANNEL COST   
>>      COUNTER      DEAD
>>      DTE            DYING       HELLO        INACTIVE LINE   
>>      LISTEN       MAXIMUM
>>      NUMBER         OWNER       POLLING      RECALL ROUTER  
>>     SERVICE      STATE
>>      TRANSMIT       TRIBUTARY   TYPE         USAGE
>>
>>     NCP>set noDE venti2:: ? ALL
>>       or one of the following:
>>      ADDRESS   CIRCUIT
>>       or one of the following:
>>      AREA          BROADCAST        BUFFER          CONSOLE    COUNTER
>>      CPU           DELAY            DIAGNOSTIC      DUMP       HARDWARE
>>      HOST          IDENTIFICATION   INACTIVITY INCOMING      LOAD
>>      MAXIMUM       OUTGOING         RETRANSMIT      ROUTING    SECONDARY
>>      SEGMENT       SERVICE          SOFTWARE        STATE     
>>     SUBADDRESSES
>>      TERTIARY      TYPE
>>
>> I may be mis-remembering how our DN200 was configured.  I believe the 
>> Top-20 NCP parser may have shared keyword tables with other DECnet 
>> implementations via BLISS at one point. Anyway, I recall at least some 
>> of it was BLISS.  I haven't look in that particular box lately, but 
>> the parse tables appear to assembler now, with heavy uses of GLXMAC 
>> macros.
>>
>> The Tops-20 operator interface has two 'sub-modes', as it were: NCP 
>> and LCP.  NCP parses DECnet related keywords.  LCP is used for LAT 
>> configuration.  The operator subsystem is kind of neat; OPR parsers 
>> send packets to Orion, a communications server and router.  It notes 
>> an NCP subset and routes that to NMLT20 which contains just what you 
>> think it would.
>>
>> Early versions of NMLT20 (about 1980) appeared insufficiently 
>> productized.  It could hang in a loop or (more frequently) crash for 
>> no readily apparent reason.  We submitted a rather large number of 
>> SPR's...  However, it's quite stable now; I don't believe I've had a 
>> single crash.  I have yet to come to a fully informed conclusion about 
>> the lexical interface; I sometimes get the impression that it parses 
>> for more than it can actually do.
>>
>> Elsewhere in Tops-20, DECnet implementation is a mixed bag; most of 
>> the monitor code appears finished; I go months without reboots and 
>> that is with active development.  The user mode code is another matter 
>> entirely.  I had my hands full fixing a number of issues with DAP and 
>> FAL.  It was /invaluable/ having heterogeneous HECnet nodes to test 
>> against, RSX 11M+ and VMS.  SETNOD needs some finishing and perhaps 
>> some additional functionality.  I have yet to look at FTS; I don't see 
>> how it can get or generate any traffic at all. The monitor interface 
>> (NFT:) does not appear to be operational.  Elsewhere (perhaps Galaxy) 
>> appears to have no submission interface.  Johnny and I have some minor 
>> tweaks to flush out with MAIL11 (in addition to the changes that I've 
>> already put in).
>>
>> On 3/2/20 3:25 PM, Robert Armstrong wrote:
>>
>>     >I don't see where I would set an id and password in 7-1-CONFIG or
>>     in NCP.
>>
>>     Don’t know about TOPS-20, but in NCP/NICE/NML it’s
>>
>>     SET CIRCUIT … VERIFICATION ENABLED
>>
>>     and
>>
>>     SET NODE … TRANSMIT PASSWORD …
>>
>>     The password the remote node sends has to match the one that’s
>>     associated with that node in the local DECnet database.
>>
>>     Bob
>>


-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol