Path: utzoo!news-server.csri.toronto.edu!rutgers!usc!samsung!uakari.primate.wisc.edu!sdd.hp.com!hp-pcd!hplabs!pyramid!pyrps5.pyramid.com!mre From: mre@pyrps5.pyramid.com (Mike Eisler) Newsgroups: comp.unix.internals Subject: Re: (was slashes, now NFS devices) Message-ID: <147986@pyramid.pyramid.com> Date: 12 Mar 91 19:02:19 GMT References: <468@appserv.Eng.Sun.COM> <147163@pyramid.pyramid.com> Sender: news@pyramid.pyramid.com Lines: 42 peter@ficc.ferranti.com (Peter da Silva) writes: >In article <147163@pyramid.pyramid.com> mre@pyrps5.pyramid.com (Mike Eisler) writes: >> Larry said "general". RFS is capable of doing Intel to Intel transactions >> as well as anything. However can OpenNet perform device accesses between >> 386's and Sparcs? >No, but it *can* do it between DOS and Unix or Xenix. And it can do it between >VAX/VMS and UNIX or Xenix. I've never run it from DOS to VMS. NFS works between all those architectures and operating systems. > (i.e., I presume the point your making is that it's not Open if it's > limited to intel. Last I heard intel didn't make VAXes. They don't Last I heard, Sun doesn't make VAXes either. But the platform they sell NFS on is a VAX. If Intel choses to limit their product to Intel chips, then no it isn't open. > necessarily make the CPUs in random DOS machines, either, but that's > a nit-pick. In any case if they can run it on DOS they should certainly > not have any difficulty with SunOS.) Cross architecture device accesses are very hard to do, because the byte ordering may be different. VAXes and Intel machines order bytes the same, so yeah, OpenNet can work. But hell will likely freeze over before you'll see OpenNet work in ALL cases between a Sparc and 386, M68k and 386, etc. Suppose one has a custom driver foo, and a custom ioctl foobar for it that moves 32 bit integers from the kernel to the application and vice-versa. Now an applicaton running on the SPARC system issues ioctl foobar to /dev/foo, where /dev/foo lives on a remote Intel server. How is OpenNet going to the byte swapping? Sun realized that a transparent hetereogenous device sharing solution was impossible, so rather than attempt to sell the world some snake oil, they settled for regular file and directory sharing. If they had done otherwise, contributors to this discussion would have blasted Sun for designing a file sharing protocol that locked out little endian architectures. -Mike Eisler mre@pyramid.com