From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 1204 invoked from network); 13 Dec 1996 20:53:57 -0000 Received: from euclid.skiles.gatech.edu (list@130.207.146.50) by coral.primenet.com.au with SMTP; 13 Dec 1996 20:53:57 -0000 Received: (from list@localhost) by euclid.skiles.gatech.edu (8.7.3/8.7.3) id PAA00444; Fri, 13 Dec 1996 15:44:48 -0500 (EST) Resent-Date: Fri, 13 Dec 1996 15:44:48 -0500 (EST) From: Zefram Message-Id: <29327.199612131942@stone.dcs.warwick.ac.uk> Subject: Re: named pipes and NFS To: coleman@math.gatech.edu (Richard Coleman) Date: Fri, 13 Dec 1996 19:42:06 +0000 (GMT) Cc: zsh-workers@math.gatech.edu In-Reply-To: <199612131852.NAA25759@redwood.skiles.gatech.edu> from "Richard Coleman" at Dec 13, 96 01:52:01 pm X-Loop: zefram@dcs.warwick.ac.uk X-Stardate: [-31]8569.10 X-US-Congress: Moronic fuckers Content-Type: text Resent-Message-ID: <"YR9I8.0.p6._0Sio"@euclid> Resent-From: zsh-workers@math.gatech.edu X-Mailing-List: archive/latest/2578 X-Loop: zsh-workers@math.gatech.edu Precedence: list Resent-Sender: zsh-workers-request@math.gatech.edu >I do not believe that name pipes/fifos (or any other >type of special file) have ever worked across NFS. The issue is not whether pipes pass data between machines (they never have). The issue is whether it is possible to create a FIFO node on an NFS-mounted filesystem -- if one can, then it should work (locally) as a pipe. In the usual case, it is possible, but Zoltan's filesystem mounted on Solaris from Linux doesn't permit it. (Which is odd; ISTR the Linux nfsd having specific support for FIFOs.) >It goes against the stateless nature of NFS. Also >remember that NFS was designed to handle more than >just Unix. But it is very dependent on Unix for its design. RFC1094 (NFS) briefly mentions the possibility of named pipes, and does explicitly support most of the other types of special file found under Unix. RFC1813 (NFS v3) explicitly supports FIFOs, and adds features like pathconf, that directly map onto Unix filesystem features. -zefram