Newsgroups: comp.parallel.pvm
From: bwh@inf.rl.ac.uk (Brian Henderson)
Subject: Re: [Q]: Messagepassing pvm ->pc
Organization: The Unconfigured xvnews people
Date: 16 Sep 1994 14:04:33 GMT
Message-ID: <35c8lh$1gdu@unixfe.rl.ac.uk>

In article ria@news.uni-paderborn.de, jam@uni-paderborn.de (Jan (not Jam) Wegner) writes:
>Dear netters,
>
>Crispin Cowan (crispin@csd.uwo.ca), a kind soul replied personaly by mail
>and wrote:
>
>|>The major result from that discussion is that no one has done it, and
>|>it would be difficult (at best) to try to do it.  The useful
>|>suggestions included:
>|>
>|>-using an OS that isn't quite so brain-dead as Windows, and can support
>|> PVM, such as Linux, NetBSD, or OS/2
>
>What about using Winsockets, or another TCP/IP-'clone', what about a tcp/ip
>connection annyway?
>
>
>jan.
>

You need more than just the ability to communicate via sockets! Having sockets
is a big help but lack of a truely multi-tasking UNIX-style system is still the
MAJOR difficulty. 

You need the underlying operating system support such that sockets can be used 
between processes on both the LOCAL and networked machines. WINDOWS sockets 
really address the network access only.

PVM uses sockets to talk to its daemon and applications. 

Starting an application on a PC under PVM requires the application connecting
to a deamon( on the local machine ) to establish links and spawn ( fork ? ) 
new processes on the local and remote machines. This is where WINDOWS lacks the
necessary support expected by PVM.
 

---
Brian W. Henderson		    *	UK JANET: bwh@uk.ac.rl.inf
Parallel Processing Group	    *	UUCP:	  ..!mcsun!ukc!rlinf!bwh
Computational Modelling Divison,    *
Informatics Department,		    *******************************************
Rutherford Appleton Laboratory,
Chilton, Didcot, Oxon. OX11 0QX,	Tel : +44 + (0235) 821900 ext. 6151
_______________________________________________________________________________



