[comp.protocols.iso] CONS question

milt@cmcvax.UUCP (Milt Roselinsky) (12/01/89)

If someone was going to run CONS over Ethernet, how does that work?  
Will there be CONS running over Ethernet?  Do you make an X.25 call to an
Ethernet address?  I don't really see how this could work, who is the
DCE/DTE, etc.  Are people really going to do this?  If not, what will
people in Europe do to connect their Ethernets to the CONS based OSI
networks?

Thanx,
Milt Roselinsky                     CMC Corp.
cmcvax!milt@hub.ucsb.edu            125 Cremona Dr.
                                    Santa Barbara, Ca. 93117

Christian.Huitema@MIRSA.INRIA.FR (Christian Huitema) (12/01/89)

Two solutions are currently used for providing CONS over Ethernets and
other LANS. One is orthodox (OIS wise) and the other one is popular.

The orthodox line is to stack the X.25 PLP (ie. iso 8208) over IEEE 802
LLC-2, and use NSAP addressing, considering that the Ethernet address
is somehow a SNPA. It is quite a piece of code, not so easy to
optimize. Five years ago, we experimented with the CNET in France a
lighter version of it (using LLC-1), but although the performances were
good, it failed to reach a general agreement in the community. A full
implementation of that has been done at Edinburgh University. It is
used in JANET, and can in principle be ordered from your favorite manufacturer.

The popular line is to stack a small convergence protocol over TCP-IP,
e.g. in order to provide packet delimitation within a byte stream. Here
again, one can use NSAP addressing, and consider the TCP-IP <host
address + port> as a 48 bits SNPA. This is implemented in ISODE, and
also in the EAN X.400 service, as well as by CISCO. Problem here is the
lack of an agreed convergence protocol (RFC?).

In both cases, a fully working solution need to use a CONS "ARP"  for
converting between NSAP and SNPAs. It is called a Connection Oriented
ES/IS protocol in the lingo, and is not completely formalized within
ISO (DP stage, as far as I remember). A secundary problem is that it is
tailored for passing X.25 addresses, which are 15 digits, and that you
need an agreement for the representation of 48 bits over 15 digits.

Christian Huitema

PS.
A full implementation of X.25 over LLC-2 has been done at Edinburgh
University. It is used in JANET, and can in principle be ordered from
your favorite manufacturer.

S.Walton@CS.UCL.AC.UK (Simon Walton) (12/05/89)

   Further to Christian's reply I would point out that DTE/DTE
X.25 connections are possible and that the method is defined in
the CCITT and ISO standards for CONS. Basically, the only
problem is that one side must act as a DCE for the purpose of
assigning logical channel numbers. This is done using restart
packets. There is no problem at layer 2, since LLC2 is
symmetric.

Simon

forster@CISCO.COM (Jim Forster) (12/05/89)

Last week Christian was discussing solutions for proving CONS over
Ethernets and other LANs.  He mentioned the orthodox way to do this, and
a popular way.

>> The popular line is to stack a small convergence protocol over TCP-IP,
>> e.g. in order to provide packet delimitation within a byte stream. Here
>> again, one can use NSAP addressing, and consider the TCP-IP <host
>> address + port> as a 48 bits SNPA. This is implemented in ISODE, and
>> also in the EAN X.400 service, as well as by CISCO. Problem here is the
>> lack of an agreed convergence protocol (RFC?).

I was a bit confused because we at cisco do not do anything in particular
with ISODE.  After discussing it with Christian, I now understand what he
meant, and think the clarification might be of interest to others.

cisco does not do anything with ISODE.  We can, however, however, switch
X.25 calls through our routers.  If the source & destination of the X.25
call are not on the same router the X.25 PLP is sent inside a TCP
connection to the remote router, across an arbitrary internet.  This
approach is similar to ISODE, in that an ISO protocol (in this case 1984
X.25, which approximates CONS) is carried in a TCP stream.  At this point
we have not done anything, such as a library, or documenting the protocol,
to allow Ethernet-based workstations to be endpoints in this scheme.


  Jim Forster
  cisco Systems