quake/articles/1997/art-1492

Path: mantis!not-for-mail

From: captain@julian.uwo.ca (Mike Kirk)

Newsgroups: rec.games.computer.quake.announce,rec.games.computer.quake.editing,rec.games.computer.quake.quake-c,rec.games.computer.quake.misc,comp.lang.java.programmer,comp.lang.java.misc

Subject: JAVA Vis -- Good Idea? Bad Idea?

Date: 10 Jul 1997 16:58:32 +0100

Organization: The University of Western Ontario, London, Ont. Canada

Lines: 27

Sender: moderate@jobstream.co.uk

Approved: quake@mantis.co.uk

Message-ID: <33b04680.374526600@newshost.uwo.ca>

Reply-To: captain@julian.uwo.ca

Excuse my ignorance. I'm just looking for off the wall opinions.

>From what I can see in the source code, I could write a Java version

of vis that either a)Uses a server section to farm off tasks to other

machines or b) chains together different parts of code using

persistent TCP/IP connections to get a sort of poor mans vector

processing.

The idea is to get more than 1 physical machine in on the act of a

VIS. What I don't know is:

1) Is the overhead of moving the data over the network not worth the

time (vs. keeping it on one machine)?

2) Taking into account the performance of Java over a wide variety of

PC's, how many seperate machines would have to be "in on the act" to

reach the critical point where it would equal 1 machine (running

native vis)? (rough estimate).

3) Would anyone use it?

I may write this anyway for practice (don't know Java network stuff

yet), but I'd like an idea of what I'm getting into before I start.

All opinions are appreciated. Thanks

P.S- if you could also send your replies via email it'd save me alot

of work. Also, yes I've posted this one once before, by all the email

on the server got clobbered so my responses are lost.