[HECnet] Cisco DECnet routers and NML

[Johnny Billquist]

On 2020-05-05 00:29, Thomas DeBellis wrote:
> I guess a number of things might be happening to build the map, then?
> 
>  1. Johnny's database gets queried for relevant data
>  2. A DECnet NICE query is sent, yet these data stores do not have
>     complete information
>  3. SNMP query

Yes, I think the idea would be something like that.

> I'm asking this for two reasons.
> 
> First, Tops-20 doesn't currently support SNMP.  However, I had actually 
> architected an SNMP server a few years ago; it would grab various bits 
> of information via GETAB%, CNFIG% and maybe Galaxy, but would have been 
> Read-Only.  Investigating a few things I had discerned from Windows and 
> some embedded devices (Linksys routers and some HP print servers) as 
> well as reviewing some RFC's, it didn't seem to be that heavy a lift.
> 
> I never implemented it because I didn't have any cool management 
> software to read all of it and display it.  I believe the only stuff I 
> was aware of were commercial offerings (perhaps Tivoli, which means 
> "Bring Your Wallet").  Maybe there is something now. Also, Orion 
> implements a remote (DECnet only) interface for most relevant management 
> (except shutdown); this could be trivially expanded to handle TCP/IP.  
> There was also the matter of security for setting items; I'm barely 
> started implementing cryptography on the 20.

I never much liked SNMP myself, even though I have worked with it in the 
past. But here it might actually be a good idea. Not that I need it for 
RSX, but with some easy arm twisting, I could be convinced to do 
something there as well.

> Second, some issues with MAIL-11 and DECnet SMTP on Tops-20 have caused 
> me to believe that one solution would be to expand the DNS service 
> (CHIVES) to handle A and MX records for DECnet.  That's not actually 
> that big a stretch; the DNS specification has 16 bit data types for 
> CHAOS, which would be trivially re-usable for Phase IV DECnet.  However, 
> I'm finding the current documentation to be a chore to wade through.  
> The source code incomplete in certain functionality.

I seriously doubt you'd ever get this to play. Not only do I think that 
CHAOS, even though it technically exists, are tested or supported by any 
name server today. I also can't really see how you think adding DNS 
would solve any problems you might have with MAIL11.

You would only be able to use DNS over Internet, while any node on 
HECnet will not be visible anywhere there, unless someone decides to 
register a domain, and then somehow populate this domain with all the 
information from HECnet, which then also needs to be maintained. But 
only your machines would ever even try to go there then to find the 
address of other machines.

And anyone externally would then have to address mails to this domain in 
order for it to reach hosts on HECnet, but you'd have to have MX records 
for each node pointing to some mail gateway or other. But internally you 
would not want to make use of these MX records, since on DECnet, you'll 
be directly talking to the end destination, unless you have an explicit 
hop path in the mail address, in which case you even less want to 
involve something like MX.

And also, anyone externally would not be trying to look for MX records 
under the CHAOS class, but still under IN, which would then fail, as the 
DECnet nodes would not have addresses in the IN class.

So it would be even more exclusively to your implementation only, and 
not usable by anyone else.

Sorry for shooting this down, but this is about as (or even less) useful 
than implementing SMTP under DECnet. You would be able to talk to 
yourself, and nobody else.

> Assuming I get this figured out, what I would do is modify SETNOD to 
> parse a fuller version of Johnny's database.  In addition to building 
> the binary file for importation into Tops-20's DECnet node name hash 
> table, it would also build a binary TBLUK% table for use by COMND% 
> instead of .CMNOD and also a zone file that CHIVES could parse for the 
> DNS records.

Good luck.

> Right now, I'm working on SETHOST, which is a client for the earlier NRT 
> protocol (Network Remote Terminal) that predates CTERM on the 36 bit 
> platform.  It's more efficient than CTERM (I have yet to compare it with 
> LAT) and will do certain things that I like that the Tops-20 
> implementation of CTERM doesn't.

Implementing where? I thought TOPS-20 as well as Tops-10 already had it. 
It's even on the RSX distribution...

   Johnny

> 
>> ------------------------------------------------------------------------
>> On 5/4/20 1:39 PM, Paul Koning wrote:
>> Nice idea.  I can certainly do this.  To make it work, the node database would have to list IP addresses for MIB access.
>>
>> The fact that the Cisco routers don't speak NICE isn't always a problem.  If they have neighbors that do I still get the information, since I will chart a circuit even if only one of the two sides tells me about it.
>>
>> If there is a chain of Cisco routers, the one in the middle of the chain wouldn't be mapped.  But, for example, an area served by a Cisco area router would be, since the mapper will try to talk to the nodes in the area and ask them what their view of the neighborhood looks like.
>>
>> If all I did was a recursive graph walk the Cisco issue would be serious, but that's not what the code actually does.  A closer approximation is a "bulk mail" query sent to every node in Johnny's database.
>>
>> 	paul
>>> ------------------------------------------------------------------------
>>> On May 4, 2020, at 1:35 AM, Tim Sneddon<tim at sneddon.id.au>  wrote:
>>> I would definitely be up for that. Maybe "hecnet-ro" for the community name?
>>> Regards, Tim.
>>>> ------------------------------------------------------------------------
>>>> On Mon, May 4, 2020 at 1:57 AM Peter Lothberg<roll at stupi.com>  wrote:
>>>> The Cisco DECnet router implementation does not speak "decnet management" as we all knew. The way we are using them the tunnel end-points are on the Internet.
>>>>
>>>> Most of the information "missing" is actually available through the SNMP MIB, so if we could agree on a common read-only community and publish the IP addresses of those routers it would be possible to complete Paul's map..


-- 
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