Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!think.com!mintaka!bloom-picayune.mit.edu!athena.mit.edu!jik From: jik@athena.mit.edu (Jonathan I. Kamens) Newsgroups: comp.unix.questions Subject: Re: named pipe across NFS Message-ID: Date: 22 Apr 91 06:45:23 GMT References: Sender: news@athena.mit.edu (News system) Organization: Massachusetts Institute of Technology Lines: 26 In-Reply-To: it1@Isis.MsState.Edu's message of 22 Apr 91 02:54:02 GMT You can't do what you're trying to do. NFS (well, at least, a good implementation of NFS :-) is smart enough to let you create a named pipe and access it as a named pipe for both read and write on the same machine. However, it does not allow you to write to the pipe on one machine and read from it on another, or vice versa. In order to understand why this is so, you have to think about named pipes as places where processes can "meet" to talk to each other. Once two processes connect to each other on a named pipe, the communication is handled in the kernel just the same as communication would be handled on a pipe created with pipe(), with local buffers in the kernel storing the data that is sent over the pipe. Why things were done this way touches upon the recent argument about whether or not NFS should support transparent access to remote devices. I would argue that it should not, and that the behavior you are observing is correct for what NFS was designed to do and what it was not. Jonathan Kamens USnail: MIT Project Athena 11 Ashford Terrace jik@Athena.MIT.EDU Allston, MA 02134 Office: 617-253-8085 Home: 617-782-0710