peter@ficc.uu.net (Peter da Silva) (09/14/88)
We need some more compute power for doing compiles and things, and thought that SV/386 running on 386 PCs would be a nice cheap way of doing things. The only problem is that we're running OpenNet, instead of some TCP/IP based thing. Would it be possible for us to either: Run some OpenNet compatible software on SV/386, Run OpenNet DOS software under Merge 386 or VPIX, or Run an extra Ethernet card in one of our 310 or 320 boxes and run TCP/IP software on that. Anyone have any ideas? -- Peter da Silva `-_-' Ferranti International Controls Corporation. "Have you hugged U your wolf today?" peter@ficc.uu.net
peter@ficc.uu.net (Peter da Silva) (09/19/88)
We need some more compute power for doing compiles, binds and so on, and thought that SV/386 running on 386 PCs would be a nice cheap way of speeding things up. The only problem is that we're running OpenNet, which uses the ISO model, instead of some TCP/IP based system. Would it be possible for us to either: Run some OpenNet compatible software on SV/386, Run OpenNet DOS software under Merge 386 or VPIX, or Run an extra Ethernet card in one of our 310 or 320 boxes and run TCP/IP software on that. Of course, the latter is most desirable... though conflicts between the OSI and DOD models might cause problems. Anyone have any ideas? As an aside, what is the history between the split between OSI and DOD?
henry@utzoo.uucp (Henry Spencer) (09/21/88)
In article <1537@ficc.uu.net> peter@ficc.uu.net (Peter da Silva) writes: >As an aside, what is the history between the split between OSI and DOD? Well, *very* concisely, the TCP/IP protocols are a bit old and have some deficiencies, but on the whole they work pretty well. ISO struck out on its own partly because of Not Invented Here syndrome and partly because anything from DoD is seen as politically unacceptable in some quarters. It is not clear that the resulting protocols have any major advantage over TCP/IP except for being ideologically cleaner (and more complex, which is a definite advantage from the viewpoint of major manufacturers: they have the resources to implement complex protocols, but those annoying little startup companies that would like to compete with them don't). The split may be moot in another decade, since neither protocol set has the performance needed for fast optical-fiber networks. -- NASA is into artificial | Henry Spencer at U of Toronto Zoology stupidity. - Jerry Pournelle | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
george@ditmela.oz (George michaelson) (09/21/88)
From article <1988Sep20.183604.2240@utzoo.uucp>, by henry@utzoo.uucp (Henry Spencer): > In article <1537@ficc.uu.net> peter@ficc.uu.net (Peter da Silva) writes: >>As an aside, what is the history between the split between OSI and DOD? > > Well, *very* concisely, the TCP/IP protocols are a bit old and have some the split predates even TCP/IP I suggest. ARPA was designed as a datagram service but the CCITT provided services in Europe were/are virtual-circuit based. IMP and the like were being taught as "nasty U.S. ideas we don't want to run" when I learned about X.25 in York. [as an experimental service, in 1980ish] When you rent a leased line, whan you run above it is probably your own buisness, but in the absence of good host-level software you'll take what the PTT provides. If PTT's had pushed datagram at their customers renting leased lines, perhaps we'd all be in the same bathtub. By the time public networks got off the ground, the split was already history. ISO/CCITT transport must be post-1982 because I recall working with ECMA specs for transport, and then seeing them massaged into the BSI/IEEE submission (to DP state I guess). At that stage, very few people in the UK had ethernet, what was available was researchy and slow and nasty. There was already in place a GEC based packet switching VC network (SERCnet later to become JANET) running X.25 and like-minded protocols on top. -this was only just starting to squeeze out the old point-to-point links from various uni's to computer centres and the like. for ISO class 4, the early documents may have had some discussion about why they used time based rather than hop based TTL. are there more important differences? Many people were either using VMS version 2.x or V7 unix, and neither come as-is with TCP/IP, but did have free/cheap X.25 solutions available. perhaps if TCP had been visible on 360's, ICL 2900's, and DEC-10's around the 70's it would have looked a better option, but when people in the UK started writing network code for their central facilities, X.25 aligned stuff came out. So a combination of lack of availability, funded nets with non DoD protocols, PTT intransigence and NIH seem to be likely. Why then did the PTT's (who tend to dominate the standards process, certainly for the CCITT and I would expect also the ISO committees outside of the US/UK) reject the datagram model? did they see an easier path to doing VC stuff, and a better initial return on their investment in H/W and S/W? I was told they opposed multiplexing at the network level because of the established charging pattern (per VC, as well as volume-per-VC) so perhaps money was the final motive, and not architectural preferences. -george -- George Michaelson, CSIRO Division of Information Technology ACSnet: G.Michaelson@ditmela.oz Phone: +61 3 347 8644 Postal: CSIRO, 55 Barry St, Carlton, Vic 3053 Oz Fax: +61 3 347 8987
ron@topaz.rutgers.edu (Ron Natalie) (09/22/88)
First your terms are clashing. TCP supports the OSI model. The MODEL says nothing about protocols, just describes the concept in layers. But on the real side, probably the best bet is to run two Ethernet interfaces. This should NOT cause a problem. TCP/IP and OSI, and just about any other properly designed protocol can coexist on the same physical ethernet. -Ron