Newsgroups: comp.parallel.pvm
From: wrankin@ee.duke.edu (William T. Rankin)
Subject: Re: PVM on Windows PCs?
Organization: Duke University EE Dept.; Durham, NC
Date: 30 Aug 1994 16:20:54 GMT
Message-ID: <33vm96$88t@news.duke.edu>

In article <CvB5q5.M3G@lehman.com>, dfields@lehman.com (Doug Fields) writes:
|> DISCLAIMER: I am not in any way talking for Lehman: these are my own opinions
|> in every respect.
|> 
|> Okay, I started this whole thread way back when. It has since degenerated into
|> the usual arguments to be expected from people who don't actually have to get
|> work done (academics) in many ways.

First point: It is never a good idea to bite the hand that feeds you.
  Asking people for advice while intentionally or unintentionally insulting
  them is not going to endear them to your efforts.

Correlary to first point: You may be suprised at how much "real" experience 
  some people in this newsgroup have :-)

Second point: After filtering out the noise that is the DOS vs. UNIX war,
  several people raised some valid questions regarding implemetation
  details on porting what amounts to a distributed application (the PVM
  system itself) that makes large assumptions on the availability of
  a multitasking OS, including the following:

     - lack of process protection.
     - lack of the concept of remote execution
     - lack of real process scheduling

The port of PVM to such a platform is a non-trivial task at *BEST*.
Many people have cited examples to support this point.

|> Let me make these points:
|> 
|>   o There are many thousands of Windows PCs installed here at Lehman
|>
|>   o There are not an insignificant number of apps which use PVM at Lehman

What have you been running these apps on in the past?

|>   o Porting the PVM apps to use a different communications protocol would
|>     be "hard" (although it is something we are considering)

To be honest, it would probably be easier than porting PVM to DOS.

|>   o Porting PVM to Windows- at least for the minor functionality we use of
|>     PVM- should be "easy"

I feel that this is a bad assumption.  Not unless you have some very high caliber
Windows *and* UNIX hacks around there.

|>   o Applications need to be made to run on Windows PCs

Personally, I would kindof worry about the prospect of expending great amounts of
energy porting distributed software to a platform that is really not designed to
support this, rather than expending the same energy perfecting the software on
a more suitable platform.

|> Sure, we could take the holier-than-thou approach and say, "You're running
|> a non-operating system, you lamer, so we're not going to give you want you
|> want."

But it is also the job of the engineer to raise the flag and say "The
specifications that you gave us are not the best solution to your problem. Here
is a better solution." 

I have found that customers often don't exactly know what they are asking for,
and need this type of guidance.

|> We'll also be working on our resumes the next day.

NC.

|> Rather than taking
|> this attitude, we decided to attempt to give the users what they wanted:
|> access to data on PVM based applications from their Windows PCs.

Okay,  then build a Windows front end that talks to a UNIX/PVM back end using
sockets.  The best of both worlds, and you don't have to reinvent the wheel.

|> As I said, there are a number of ways of doing this.
|> All I wanted to know was if anyone had done any significant work on 
|> porting PVM to Windows PCs.

And as many people have tried to point out, this is a non-trivial task,
and perhaps you efforts would be better spent elsewhere.

Have you looked at running a minimal Linux system as a "PVM server" on each PC?
If you really strip down a Linux system, you can fit it into less disk space
than most Windows applications. :-)

|> I suppose that in the future I will stay away from asking potentially
|> religious questions. Indeed, my first reaction was simliar to many of yours:
|> "Why would you want to do that?" My job, however, is merely to get it done,
|> and come hell or high water I'm going to do it.

My sympathies ;-), and best of luck.

|> Doug

-bill

----                                /       __/    /    /
bill rankin                        /              /    /
wrankin@ee.duke.edu               ___  /    /    /    /
philosopher/coffee-drinker       /    /    /    /    /
                                /    /    /    /    /
                             _______/  __/  __/  __/



