sjl@amdahl.UUCP (Steve Langdon) (08/28/85)
I started this sequence of messages by asking a question for a friend. Little did I expect that it would spill over into an area that I work on directly. Conversations about datagrams versus virtual circuits usually generate much more heat than light, so I will have to be careful about what I say. First lets deal with what Phil R. Karn had to say: >The refusal of any of the commercial standards organizations to consider >datagram protocols is what prompted the DoD to develop their own. While I have significant personal experience with the difficulties of getting connectionless protocols through standards bodies I do not think that the DoD's development of TCP/IP had anything to do with problems in national or international standards bodies. If you look at the timing, these protocols were being developed before there was any major fight in standards groups. In fact ex-Arpanauts like Larry Roberts had a lot to do with the development of X.25 and they were probably responsible for their fair share of the problems. One of the sadder aspects of the whole debate in the standards group has been the attitude of many of the best known characters in the Arpa community. They have almost all felt that the grubby business of achieving political and technical compromises was beneath their dignity. The fight to add connectionless protocols to the OSI reference model was not fought by the people working on TCP/IP. Rather was it fought by people working for computer companies who wanted to be able to develop standards based networks that would meet the needs of their customers. The government, primarily represented by the NBS (and their contractors, especially BBN) did a large amount of work, but the Arpanet community did almost none. Why did people like Lyman Chapin (Data General), Dave Piscitello (Burroughs), and Dave Oran (DEC) work so hard to get connectionless protocols accepted in the international standards community? Because they believe, as I do, that connectionless protocols are important for us to build the networks our customers need. In particular, connectionless network protocols are the best solution to internetworking, especially when a variety of different subnetwork technologies are involved. Now lets get back to other people's comments. Mark Horton started by saying: >As I read the intent of the CCITT folks working on OSI (and this is >based on an OMNICOM seminar taught by Hal Folts, so his biases may >be what you're seeing here) they think virtual circuits are wonderful >and that anyone who would do anything with a datagram must be crazy. >As evidence to support this, they point to the fact that in 1980, >X.25 had a datagram facility, but by 1984 nobody had implemented it, >so it was deleted from the 1984 spec. (It seems to me that there was >no serious demand for datagrams from the common carriers until around >1982 or 1983 when TCP/IP became popular - CSNET would have been the >first real user.) Mark's parenthetical comment shows sensible caution about the source of the information. Hal's world view is CCITT oriented, and it shows. It is useful to remember that the CCITT did not develop the OSI model. I would suggest that anyone who discusses OSI in terms of CCITT rather than ISO has provided a very valuable piece of information about their natural biases. Now lets look at the importance of the X.25 datagram story. First you should look at what people do with X.25 networks. They use them to connect terminals to computer systems. We all know that this is not where computer networking is going, but the traffic that pays the bills is terminal to host. Since the X.25 networks were (and are) used primarily as an alternative to ordinary dial up terminal access, there was not much demand for networking sophistication. The most important technical activity (in terms of the revenue it generated) was to define new PAD type protocols so that customers could use their bisync terminals over the X.25 networks. In this environment it is not surprising that the network vendors did not spend a lot of time implementing the datagram option. In fact it took most of them years to get the 1980 version of X.25 implemented. While this was the situation in the X.25 area, work was going on in a variety of places to develop networks that supported computer-to-computer communication in a more general way. On the Arpanet, at Xerox, at DEC, etc. Most of these schemes involved some sort of internetworking. They all ended up with an architecture that had a connectionless network layer and a connection oriented transport protocol (some also used connectionless higher layer protocols, but thats another story). Now back to Mark. >Most of the interest in TC4 and CLNS seems to be from the people doing >the AUTOFACT stuff - mostly American computer companies. The Telcos >which make up CCITT have decreed that there will be no wide area >network protocols except X.25, and they are trying to move the LAN >protocols in the connection oriented direction. (Did you know that >IEEE 802.3 has a connection oriented mode? Does anyone care?) They >have gone so far as to get DOD to announce that the preferred interface >to the ARPANET is no longer 1822, it's X.25. I have no idea how you are >supposed to get through a virtual circuit oriented X.25 bottleneck to >send IP packets out over the ARPANET, this must make an interesting story. As he says, the main proponents of Class 4 Transport and the IP-like Connectionless Network Protocol, are computer companies. The main reason is that customers like to buy equipment that works. Networks of the type people want to buy today are mixtures of LANs, packet switched WANs, and leased lines (it also helps if you can wave your hands about voice and video). These networks can be built using the TC4/CLNP approach. X.25 may be overkill for some forms of WAN connection, but one of the virtues of the CLNP is that it will happily run over almost any subnetwork. [As an aside, Amdahl builds X.25 packet switches and PADS so I am certainly not against X.25. X.25 has a very useful place in the overall network, but that does not make it the correct internetwork protocol to use for OSI.] The other side is the PT&Ts who want to extend their influence into the computer room. Until recently most of them would not have recognized a LAN or a higher layer protocol if it bit them, so they are naturally taking the approach of extending the protocols they know to cover the new territory. Unfortunately they are doing it by dictating standards without any field experience. Given their political clout, they are having some success at achieving this goal, especially in Europe. The main reason for optimism is the work going on in the OSI Implementors' Workshops and associated efforts (MAP, OSINET, TOP, etc.). A large number of companies have agreed to implement a common stack of OSI protocols (I can post details if people are interested). If we can keep this momentum up, we may get working networks based on real standards. If we fail, then the PT&Ts will get to find out how well their proposals stack up against one large computer company's proprietary architecture. -- Stephen J. Langdon ...!{ihnp4,cbosgd,hplabs,sun}!amdahl!sjl [ The article above is not an official statement from any organization in the known universe. ]
sjl@amdahl.UUCP (Steve Langdon) (08/30/85)
In my recent posting on communication standards <1953@amdahl.UUCP> I mentioned that: >The main reason for optimism is the work going on in the OSI Implementors' >Workshops and associated efforts (MAP, OSINET, TOP, etc.). A large >number of companies have agreed to implement a common stack of OSI protocols >(I can post details if people are interested). If we can keep this >momentum up, we may get working networks based on real standards. After watching the continued discussion I decide that it would be a good idea to post the details I mentioned. Mark Horton referred to one of the pieces when he mention AUTOFACT '85. The other piece is the OSINET effort that the NBS are organizing. AUTOFACT is a trade show happening this November in Detroit. At the show over 20 companies will demonstrate a working internetwork based on OSI protocols. The network will include IEEE 802.3, IEEE 802.4 (5 and 10 Mbit), and X.25 subnets all connected together by routers. Protocols used include those implied by the subnets (IEEE 802.2 LLC type 1 class 1, and X.25) and 1) CLNP (the OSI version of IP) 2) Transport Class 4 3) The OSI Session kernel 4) File Transfer, Access and Management (FTAM) OSINET is an NBS organized, long term network for testing OSI protocol implementations. The participants (13 at last count - Amdahl is a participant) will use a number of X.25 networks as a wide area backbone to allow ongoing protocol testing and to exchange information about OSI. The initial protocol suite is the same as that used at AUTOFAC, but the goal is to add other OSI protocols to the mix. While these efforts do not represent a large running internetwork like the Arpa Internet, they go a long way towards proving that the OSI protocols are ready to make their way into products. -- Stephen J. Langdon ...!{ihnp4,cbosgd,hplabs,sun}!amdahl!sjl [ The article above is not an official statement from any organization in the known universe. ]