dbr@foxvax5.UUCP (01/29/86)
------ This is a request for ideas from those of you who have been through this before. We are currently implementing GM-MAPs protocol standard in the UNIX(tm) domain. We would like to do so under the 4.2 "socket" front end, but there are a number of issues that seem to imply that such a layering cannot be done (I am not completely familiar with the MAP standard, so if I get it wrong don't jump too heavily): 1. Connections are one to one with processes. What this implies is that children can NOT inherit an open connection from a parent. This will break many, if not all, of the listener/server programs written in the socket domain. 2. Assume a executable image (a.out) is "registered", but NOT running. A new UNIX process must then be created when a "connect" call is issued. How? A server is the obvious solution (wait for MAPs level requests to start processes), but how can the arguments and path be passed? The "yet another server" approach is ugly... There seem to be many of you out there who have already implemented the MAPs standard given the article in IEEE news a few months back. Please let me know your thoughts on the matters above - and on any others that I haven't run into yet (but you know I will...), or point me to any previous mail/discussions of GM-MAP and UNIX. (We only get limited newsgroups here, so if you are going to point me there, please send an archive if possible.) Thanks. Douglas Robinson The Foxboro Company Department 350 38 Neponset Avenue cybvax0!foxvax5!dbr Foxboro, MA 02035 617/543-8750