[mod.protocols.tcp-ip] VMS TCP/IP

ron@BRL.ARPA (Ron Natalie) (08/25/86)

I have been trying to help our VMS people get their TCP/IP working.
What we are using is the DTI->COMPION->GOULD->INTERNET VMS code now
called Hyperlink.  I spoke with someone who claimed to be the person
working on the code and they seem to be a little naive about how IP
works.

The problem is to set up the routing table it is required to make
entries for every gateway in the internet.  They have a program that
generates this from a NIC host table, but here lies the problem.  There
is no concept of DEFAULT route, nor is EGP used so there is no way to
make this system work reliably in todays Internet.  Consider the following
network.

	VMS HOST <--NET A--> GATEWAY <--NET B--> GATEWAY <--NET C--> TARGET
	addr A.v           A.g     B.g         B.h      C.h        C.t

There is a VMS machine on net A with address A.v, a gateway between nets
A and B with interface address A.g and B.g respectively, etc...
To get the Hyperlink code to route to the TARGET, you need the
following enties in their table

	Interet Address		Gateway
		A.g		B.g
		B.g		A.g
		B.h		C.h
		C.h		B.h

Note there are four entries here (don't ask me why it needs them in
both directions).

The problem comes in that gateways must be up to date in the nic-table.
Note that this neither provides for the dynamics provided by EGP, nor
does it help people who are not on MILNET but on little stubs that the
entire world is on the far side of a single gateway (such as the VMS
machine at BRL).

When I mentioned this to them, they said no one had ever complained
before, and that the X.25 validated system solved the problem with
the name daemon (right).  I don't believe that either.  Anyway, since
they seem to be ignorant of the problem, I guess I'm going to half to
kludge up a table that lists every NET number in the net and fake it
pointing at the DEFAULT gateway and vice versa.  It turns my stomach.

It's really amazing what you can get away with in a DDN-validated system.
Just what are they validating?  By the way, we can't use Woolengong since
the refuse to provide any way for us to handle IMPs on Class B internet
number networks (like either doing it for us, or providing some way to
get the source).  And as far as VMS is concerned, I think it is a shame
to leave a VAX running diagnostics all day.

-Ron