Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!uunet!seismo!rutgers.rutgers.edu!princeton!udel!mmdf From: ncoverby%NDSUVAX.BITNET@wiscvm.wisc.edu Newsgroups: comp.os.minix Subject: (none) Message-ID: <365@louie.udel.EDU> Date: Mon, 20-Jul-87 13:12:41 EDT Article-I.D.: louie.365 Posted: Mon Jul 20 13:12:41 1987 Date-Received: Tue, 21-Jul-87 05:19:09 EDT Sender: mmdf@udel.EDU Lines: 30 From: Glen OverbySubject: Fixing the tty driver Last night I applied Andy Valencia's serial port driver to Minix 1.2. Looking at this implementation, which seems to follow Tanenbaum's ideas, gives me the impression that it is essentially a fall-back to a monolithic system. The way I propose to clean up the TTY driver is to convert the TTY task to a streams system. The Unix streams system described by Dennis Ritche[1] uses message passing between streams processors. Since Minix already has message passing, streams could be implemented as a set of server processes. There are a few problems with this. First, Minix's messages do not queue. There will have to be either some sort of queueing within the streams processors or something. Streams will also increase the number of messages being passed. What ever happened to mpx (multiplexed) files? I would think that they would be ideal for windowing systems (or at least mpx devices). Opinions? References: [1] "A Stream Input-Output System" D. M. Ritchie BSTJ V63 No. 8 (Oct '84). ----------- Glen Overby Bitnet: ncoverby@ndsuvax from UUCP: ihnp4!psuvax1!ndsuvax.bitnet!ncoverby