Path: utzoo!attcan!utgpu!news-server.csri.toronto.edu!rutgers!uwm.edu!wuarchive!emory!stiatl!tok From: tok@stiatl.UUCP (Terry Kane) Newsgroups: comp.os.os2 Subject: Re: Distributed Processing/Broadcasting Message-ID: <10268@stiatl.UUCP> Date: 2 Jul 90 17:48:17 GMT References: <1990Jul1.023719.29550@murdoch.acc.Virginia.EDU> Distribution: usa Organization: Sales Technologies Inc. Atlanta, Ga. Lines: 35 tg2r@dale.acc.Virginia.EDU (Todd Greenwald) writes: >I'm developing a multiuser distributed processing application based >on a client-server model. When one client alters data, the server >is notified and then 'broadcasts' the change to all other registered >clients. IMHO, your problem is more basic than the mechanics of broadcasting - your problem lies in choosing to broadcast IFF you are using a high speed net e.g. >5mbps, or even 1mbps for that matter. I have worked on a product for air traffic controllers which used multipoint serial communications, but by george (btw, where was he?) I would have gone with secure point- to-point circuits if I'd had more bandwidth! I would want to know who the user is - the tower? ground control? or terminal personnel/passengers? That makes all the difference. If this were for ATC usage, I'd certainly want to see this on a dedicated hi-speed net with redundancy. But then you should also bear in mind that the FAA requires Unix as the platform for software put out to bid. If the user community was not mission critical, I certainly wouldn't worry about the single-cast comms overhead, even at low speed, and I would go with single-cast because that cuts the complexity of your work by at least an order of magnitude, and hey, like my daddy always said - if you can get 90 percent done for 10 percent work, go for it! I suppose that this is slightly out of the realm of this newsgroup, but, I felt that I'd share my experience with you. standard disclaimer -- Terry Kane gatech!stiatl!tok Sales Technologies, Inc 3399 Peachtree Rd, NE Atlanta, GA (404) 841-4000