jwest@aplcen.apl.jhu.edu (3619) (11/22/89)
I have some questions that I hope some people on the net can help me with. I am currently working at NIST and examining the problem of inter-connecting of OSI Connection-mode and Connectionless networks. Ultimately we plan to produce a paper outlining the possible solutions to this problem (outlined below). Here's the motivation: ---------------------- The United States seems to be moving towards and architecture where hosts communicate using Transport class 4 over CLNP (I know that the U.S. Gosip requires TP-4 over CLNP), while the Europeans seems to be moving towards an architecture where hosts communicate using Transport class 0 using X.25 to connect to connection oriented networks. Not only are CONS and CLNS incompatible, but so are the TP-4 and TP-0 protocols. Here's the problem: ------------------- How do you inter-connect these networks (with out requiring a change in either North America or Europe)? Also, considering the large TCP/IP user base in the U.S. (and the one growing in Europe) how would you connect TCP/IP networks in with the two types of OSI networks. I know Marshall Rose has proposed a way to run TP-0 on top of TCP, so that problem may have been solved (at one end). What we need is some type of general relay to connect the two (three?) different types of networks. A little Discussion: -------------------- Here is the basic topology: (N is less than 10) (Lets call this a BLOCK) ========= --->| Relay |<- +---------------+ / | #1 | \ +---------------+ | CNLS | / ========= \>| CONS | | Network |< . | Network | | | . | | | | . | | | |< ========= ->| | +---------------+ \ | Relay | / +---------------+ ^ ^ --->| #N |<- ^ ^ | | ========= | | HOST#1 ... HOST#K HOST-A ... HOST-? Now we don't want to stop here, We need to concatenate these units together to form an internet: BLOCK-1 <---> Relays <---> BLOCK-2 <---> ... <---> BLOCK-M (Most likely more than 1 relay) The OSI reference model demands that these relays operate at the Network layer, yet this creates problems (according to ISO/TC97/SC6/WG2 Dec 1987 document "NLR Approach to CO-CL Interworking) with routing (+ other problems) , and there is always the problem of incompatible transport protocols that must be converted. The relays could operate at the Application level; I believe that there already exist such relays for X.400 <--> TCP/IP mail (I forget the RFC#). But what if someone wanted to use FTAM? If a solution was developed at the Application layer, performance would suffer. So what about a solution at the Transport level? This would violate the one of the principles of the OSI reference model, but there might be some performance gains and the ability to bypass some of the problems at the Network layer (for a new set of problems at the Transport layer ;-) ). My Questions: ------------- * Have I missed anything in my above discussion? * Is anyone familiar with relays that that operate at the Transport layer for either: - connecting networks that provide OSI COnnection-mode Network Services to OSI ConnectionLess Network Services (CLNP), - or connecting OSI networks to TCP/IP networks. and can provide some references (or comments). * Are there any currently implemented solutions to this problem? (At any level) * Anyone have an opinion on at what level in the OSI stack a relay should be constructed? Why? If there is interest I will summarize any results to the net. Thanks in advance for your help, Jim West National Institute of Standards and Technology Gaithersburg MD., 20899 (301) 975-3619 west@osi.ncsl.nist.gov jwest@aplcen.jhu.edu
Stef@NRTC.NORTHROP.COM (Einar Stefferud) (11/24/89)
In-reply-to: jwest@aplcen.UUCP (3619) message of 21 Nov 89 16:41:34 +0000. <3999@aplcen.apl.jhu.edu> > I am currently working at NIST and examining the problem of > inter-connecting of OSI Connection-mode and Connectionless networks. > Ultimately we plan to produce a paper outlining the possible solutions > to this problem (outlined below). I will not bother you all with another copy of the problem description! TS-Bridging has been used for a year now to solve all of the problems referred to, without any loss of application layer functionality. At least 3 implementations are in regular use today, and one of them is freely available as part of the ISODE. I should also mention that The Open Book spends nearly 100 pages looking at all the various options when going between TCP/IP and OSI, including several additional options not mentioned by Jim West. I suspect that your report would do well to include all this material by reference. If you do not yet have a copy of The Open Book, I am sure we can find one in your local technical book store. Best...\Stef