[HECnet] DECnet area router configuration

[Johnny Billquist]

RSX have taken another step that I don't know if you are aware of, Paul.

Yes, NCP is for the volatile database only. RSX then also have CFE,
which works pretty much the same way as NCP, but for the permanent
database. And as a third tool, there is VNP, which is odd in this sense,
since it affects parts of the DECnet operations, but on the image that
is booted (that is, you can perform a bunch of NCP commands on the
volatile database, but it's kind of "permanent" since it's done on a
file image that is booted, so in effect you are doing the NCP commands
before the system even boots).

However, RSX also moved the nodename database totally out of the main
DECned code, and have in fact a separate task that actually handles
nodename to number translations, called NNS.

So, both SCP and NCP as well as all tasks indirectly talks with NNS
whenever a nodename is given.

NCP under RSX (and CFE and VNP) furthermore don't have a COPY command.
But in connection with NNS is also a task called NNC, which can be used
to pull node names from another node into the nodename database used by
NNS. So that do the same thing as COPY in VMS (and some more stuff).

I suspect that one reason NCP and CFE was two different tasks in RSX was
probably for resource reasons. It was probably difficult to combine them
into one program for small RSX systems.

	Johnny

Paul Koning wrote:
Ditto for the other DECnets.   "COPY KNOWN NODES" is a VMS-specific
extension.
RSX is a particularly odd case, with its separate utility for handling
volatile vs. non-volatile stuff.   That's not what the architecture
called for.   DECnet/VMS and DECnet/E are much closer to standard.   Not
that the network management spec was ever a complete standard; it was
inevitable that every OS would need OS-specific extensions, the question
was only how closely the final result would resemble what the spec talks
about.
	paul 
-----Original Message-----
From: owner-hecnet at Update.UU.SE [mailto:owner-hecnet at Update.UU.SE] On
Behalf Of Johnny Billquist
Sent: Thursday, December 03, 2009 4:03 PM
To: hecnet at Update.UU.SE
Subject: Re: [HECnet] DECnet area router configuration

Well, the procedure in RSX is totally different anyway, so that's no
help. RSX have a totally separate task to handle node name stuff.
And NCP is only used for the volatile database.

	Johnny

Steve Davidson wrote:
You need to use the "with purge" option for this to happen - at
least
in
VMS anyway.

-Steve

-----Original Message-----
From: owner-hecnet at Update.UU.SE [mailto:owner-hecnet at Update.UU.SE]
On
Behalf Of Johnny Billquist
Sent: Thursday, December 03, 2009 15:21
To: hecnet at Update.UU.SE
Subject: Re: [HECnet] DECnet area router configuration

Actually, I'm not entirely sure (I don't use VMS much nowadays), but
I
think it might remove all previous definitions before doing the
copy.
But that is easy to test. Just add a definition for some odd node
that
don't exist, and then do a copy.

	Johnny

Ian McLaughlin wrote:
Does a COPY KNOWN NODES FROM xxx remove nodes in your local
database
that aren't listed any more?

It's fairly easy to run a COPY KNOWN NODES command once in a while.
I
guess the only piece missing is an automated way for MIM to get
updates
from everyone else.

Ian.

On 2009-12-03, at 12:10 PM, Sampsa Laine wrote:

The only (and this is a very minor) benefit that I can see in a
distributed naming system is that this way each owner of say an
area
could update the name database for his network and have it
automagically propagate, rather than a centralised system we have
right now which requires your time to keep up to date.

But it's not really that big a benefit to warrant the effort -
just
automate the periodic copying of the database from MIM would be my
suggestion as well...

Sampsa


On 3 Dec 2009, at 20:07, Johnny Billquist wrote:

Sridhar Ayengar wrote:
Johnny Billquist wrote:
Node list don't even get transmitted within the same area.
Node names are local to each machine, and it is perfectly valid
to
have
different names for the same node number on different machines
(although
perhaps confusing).
How hard would it be to write software as equivalent to DNS?
Not
necessarily for general-purpose use, but just for HECnet?
For what? Just copying the nodename database between machines?
The
software can already do that, so it would just be a question of
automating it a bit.

If you'd like to get a name lookup done from some central place
at
each
nodename lookup would be almost impossible. You'd need the source
code
for DECnet, and the ability to recompile it for that to be
possible.
Not
likely, I'm afraid.

Johnny

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

---
Filter service subscribers can train this email as spam or
not-spam
here:

http://my.email-as.net/spamham/cgi-
bin/learn.pl?messageid=E4AFF6EEE04711
DE98D9899E93ED0201



--
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


-- 
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