paulb@mlacus.oz (Paul Bandler) (04/10/90)
Here are the replies I received to a posting about I placed asking for
information on running OSI applications over TCP/IP networks. The replies
vary from - oh no you can't do that, to yes you can do that, its been done
and is specified by RFC1006.
I'm still interested to here more info on the subject - particularly re: the
current status of RFC1006, products that implement it, and the likely market
for products that can work in such networks.
Here are the replies:
From: Erik Skovgaard <eskovgaa%uvcw.uvic.ca@munnari.oz>
To: paulb@mlacus.oz
In-Reply-To: <9003290920.28824@munnari.oz.au>
Message-Id: <35*eskovgaa@uvcw.UVic.ca>
Status: RO
As long as the implementations adhere to the same protocol specifications
you should not have a problem.
The question is: do they?
This is the subject of much intensive work at the moment. Several
comapnies have produced "conformance testers" that vendors are
supposed to take their implementations to before releasing the
products on the market. In reality, however, you don't know for sure
until you have tried (this is also called an interoperability test)
It does not matter that the Upper OSI layers run on top of OSI Lower
Layers or TCP/IP, as long as you have the same level of Service from
the lower stack. As it turns out, TCP/IP provides a service that is
quite comparable, so it is possible.
Of course, you must run the OSI Upper Layers at both ends. It does not
buy you interoperability with TCP/IP applications. For that you need a
gateway.
....Erik.
From: Andrew Worsley <A.Worsley%cs.ucl.ac.uk@munnari.oz>
Status: RO
To: comp-protocols-iso-x400@uunet.uu.net@munnari.oz
Path: mlacus!paulb
From: paulb@mlacus.oz (Paul Bandler)
Newsgroups: comp.protocols.iso.dev-environ,comp.protocols.iso,comp.protocols.is
o.x400,comp.protocols.tcp-ip
Subject: OSI Over TCP Interoperability Issues - Q's
Keywords: OSI TCP
Message-ID: <404@mlacus.oz>
Date: 29 Mar 90 05:33:40 GMT
Organization: The Australian Centre for Unisys Software
Lines: 29
....
However, from what I've read I'm still unclear as to the actual interoperabilit
y
that can be achieved without actually having the same implementation at both
ends.
If I take one implementation of OSI layers 5-7 stack running over TCP
will it inter-operate with such a stack of a different implementation?
Absolutely! I have implemented an FTAM completely seperately from ISODE
and did some interworking with ISODE over the TCP protocol. Also some
interworking tests with DEC FTAM and ISODE over X.25. The OSI over TCP is
exactly TP0 over TCP, all interworking problems found were to do with real
problems in the implementations at the higher levels. If you know TP0 you
will know that there is very little to it and so very little that can go
wrong. It is defined in RFC 1006 (unless another RFC has obsoleted that one)
and the only possible difference I know is that it allows you to use TPDUs of
64k not just the pidling 2k limit that TP0 (use to?) impose. This is just
an efficiency measure (to make the best use of ethernets and the like) and
I haven't heard of anything that has problems changing down to the real sizes
but I'm sure some one could make problems if they tried to.
What are the issues here? I've seen discussion on:
Transport Disconnect protocol differences
Address Format Differences
I don't see how these can be different. The exact TPDU as would be sent over
the wire via a normal OSI network connection is sent. I can't think of
anything about the sequent of activities which might be illegal or non
conformant, but I only implemented it, I don't claim to understand OSI....
"Directories issues"
I think this is a different issue altogether.
but I was unable to discern an agreed position of these issues. Can anyone
offer what the state of the art is? I've also heard mention of a standard
which defines how OSI should use TCP. Can anyone point me towards this?
The RFC 1006 is obtainable from CSIRO DIT, Australia (phone 347 8644 ask for
Andrew Waugh) if this is what you want.
....
Andrew Worsley
formerly of CSIRO DIT
id AA17905; Fri, 30 Mar 90 07:52:07 EST
Date: Fri, 30 Mar 90 07:52:07 EST
From: rob%cos.com@munnari.oz (Rob Clark)
Message-Id: <9003301252.AA17905@cos.com>
To: paulb@mlacus.oz
Subject: Re: OSI Over TCP Interoperability Issues - Q's
In-Reply-To: your article <404@mlacus.oz>
News-Path: uunet!cs.utexas.edu!yale!eagle!mlacus!paulb
Status: RO
> I've observed discussion on the net in the area of running
> OSI upper layers over TCP lower layers and visa versa. I'm also aware of
> the ISODE work.
>
> However, from what I've read I'm still unclear as to the actual interoperability
> that can be achieved without actually having the same implementation at both
> ends.
>
> If I take one implementation of OSI layers 5-7 stack running over TCP
> will it inter-operate with such a stack of a different implementation?
>
I would presume so, providing it is the "same stack", but I also believe
you are getting into deep water when you start mix & matching stacks
between ISO and TCP - if this is what you are driving at.
> What are the issues here? I've seen discussion on:
>
> Transport Disconnect protocol differences
> Address Format Differences
> "Directories issues"
>
> but I was unable to discern an agreed position of these issues. Can anyone
> offer what the state of the art is? I've also heard mention of a standard
> which defines how OSI should use TCP. Can anyone point me towards this?
>
TCP won't interwork with OSI Transport. Period. I think the postings you
were referring to were concerned with the service definitions being different.
OSI addressing is not nice even within a pure-OSI environment, let alone
trying to interwork with a TCP/IP style system.
The OSI directory is still undergoing development - who's in charge
of the phone book if you like!
Hope this helps. If not, don't hesitate to let me know!
I spend most of my life trying to get OSI implementations to interwork
and that's hard enough without throwing TCP into the works :-)
> Paul Bandler
> ACUS - Australian Centre For Unisys Software
-Rob-
-- Rob Clark, Corp. for Open Systems, McLean, VA
-- (really from:
OSI Test Development, National Computing Centre, Manchester, UK)
-- rob@cos.com, ..!uunet!cos!rob
-- Rob @ Akbar and Jeff's Protocol Testing Hut
"Where the Elite meet to complete the Test Suite"
From A.Worsley%cs.ucl.ac.uk@munnari.oz Tue Apr 3 20:21:45 1990
Received: by murtoa.cs.mu.OZ.AU (5.61+IDA+MU)
id AA11388; Tue, 3 Apr 1990 20:21:45 +1000
(from A.Worsley@cs.ucl.ac.uk for paulb@mlacus.oz)
Date: 3 Apr 90 10:19
From: Andrew Worsley <A.Worsley%cs.ucl.ac.uk@munnari.oz>
To: paulb@mlacus.oz
In-Reply-To: <Your message of Mon, 02 Apr 90 20:10:36 +1000. <9004030325.13733@munnari.oz.au>>
Message-Id: <663.639134485@UK.AC.UCL.CS>
Status: RO
Subject: OSI over TCP Nets
To: A.Worsley%cs.ucl.ac.uk@munnari.oz
Date: Mon, 2 Apr 90 20:10:34 EST
From: Paul Bandler <paulb@mlacus>
X-Mailer: ELM [version 2.2 PL0]
TP0 over TCP approach for interworking, and you mentioned that you'd
experienced interworking ISODE with DEC FTAM. Do you know whether these
implementations set out with the explicit intent to use the RFC1006 approach?
(is it a product feature of DEC OSI and ISODE to conform to conform to RFC1006?)
I think I have misled you here. I did the interworking using TP0 over X.25.
The other site was DEC which was not connected to me via any TCP link,
I didn't even think to ask if they used TCP I just assumed I get back a
blank stare. Things may be different now. ISODE certainly supports RFC1006
as the principle code writer, Marshall T. Rose wrote RFC1006.
What I'm driving at here is that I know that a number of OSI upper layer
products in my company can be transported over TCP networks (to talk to
themselves) but I'm unaware of whether they have used any 'standard' technique
to do so. If they haven't explicity implemented RFC1006, will they be able
to talk to any other implementation over TCP do you think? Maybe I just need
to ask them if they've ever heard of RFC1006.
I think you better ask them. I believe that the Internet is intending
to go to OSI protocols, running them over RFC1006 as an intermeadiate
step. Certainly NSFnet supports RFC1006 so there (is?) will be a
significant user base in the U.S. using RFC1006 it is worth doing. If
you have a TCP library it is trivial to add if you do TP0 over X.25
unless you have a very funny X.25/TCP interface.
From: Lyle Seaman <samsung!sununix.comm.wang.com!lws@munnari.oz>
Message-Id: <9004021726.AA13533@sununix.comm.wang.com>
To: paulb@mlacus.oz
Subject: Re: OSI Over TCP Interoperability Issues - Q's
Newsgroups: comp.protocols.iso.dev-environ,comp.protocols.iso,comp.protocols.iso.x400,comp.protocols.tcp-ip
References: <404@mlacus.oz>
Status: RO
In comp.protocols.iso.dev-environ you write:
>If I take one implementation of OSI layers 5-7 stack running over TCP
>will it inter-operate with such a stack of a different implementation?
WELLll, to be precise, the ISODE is an implementation of OSI layers
4-7, running over TCP. It uses TCP as a Network Service Provider.
>offer what the state of the art is? I've also heard mention of a standard
>which defines how OSI should use TCP. Can anyone point me towards this?
I think you're thinking of RFC 1006, I don't know if it has become an
official standard.
--
Lyle Wang lws@comm.wang.com
508 967 2322 Lowell, MA, USA uunet!comm.wang.com!lws