Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!sdd.hp.com!elroy.jpl.nasa.gov!decwrl!nsc!pyramid!athertn!joshua From: joshua@athertn.Atherton.COM (Flame Bait) Newsgroups: comp.sys.apollo Subject: Re: OSF Picks Apollo's NCS for Distributed Computing Message-ID: <24920@joshua.athertn.Atherton.COM> Date: 7 Jun 90 20:10:09 GMT References: <1990Jun1.180211.3586@idacom.uucp> <1056@fiver.UUCP> Reply-To: joshua@Atherton.COM (Flame Bait) Organization: Atherton Technology, Sunnyvale, CA Lines: 33 > = palowoda@fiver.UUCP (Bob Palowoda) # = someone else quoting an industry rag > My question is can both of them [Sun's and Apollo's RPC systems] coexsist > on the same hardware? Yes. I've done it, and so have you, if you run NFS while using Apollo's RPC. NFS is almost always built on Sun's RPC. It is for exactly this reason that I believe that standarizing RPC systems is like standarizing languages: a good think for a small company to do, an OK thing for a large company to do, but a bad think for a bunch of vendors to do. What would people say if OSF announced that C was the standard language for programming. FORTRAN, Pascal, and AWK would not be encouraged... Like programming languages, different RPC protocols have different strengths and weaknesses. Different problems will require different RPC systems; one size does not fit all. I do not see what the gain is in having OSF (or UI) choose one RPC system as "standard". Why not support all the common RPC systems? There are only 2 or 3, they each have special strengths and weaknesses, and do not interfear with each other. >Question, what is the best attribute of NCS that will make NFS users >change over to it? I'm not sure that this is the goal. I think that OSF wants new applications to be written using NCS. Furthermore, it may not be possible to implement NFS over NCS. The NFS spec requires that servers be stateless, it may not be possible to do this if your RPC system is storing state. (That is the "official word" on why NFS uses Sun RPC/UDP, instead of TCP.) I'm not saying that stateless file systems are a good thing, either. Joshua Levy (joshua@atherton.com)