Bolot,
     
     Thanks for your suggestions.
     
     The application is actually a robotics application which I can't say 
     much about because of nondisclosures in my contract. The contract 
     actually requires C++ components but I am planning on doing a Squeak 
     version simultaneously. Hopefully, I can sometime convince them of the 
     superiority of Smalltalk :-)
     
     Basically, the problem is to be able to interact with a robot and 
     associated components (processes) either locally on a non-networked 
     machine or via the Internet. In either case, there will be several 
     separate communicating processes involved.
     
     I am really happy about your interest in creating support for 
     distributed applications. 
     
     Something which I mentioned previously in the list was a desire for 
     generic support for distributing an application on the fly as opposed 
     to making sure up front that the necessary classes are available on 
     the distributed images. In this case, an application might determine 
     that it made sense to distribute itself and would look for running 
     images on the network. If the available images did not have the 
     necessary classes available, it would be nice to have generic support 
     for their creation, use, and appropriate rerouting of messages from 
     the local classes. Generic support for ensuring permanence or removal 
     of the distributed classes would also be nice. I guess the other thing 
     is generic support for finding and connecting to the other running 
     images.
     
     Jerome


______________________________ Reply Separator _________________________________
Subject: Re: Naive question - Communicating between multiple images
Author:  Bolot Kerimbaev <bolot at cc.gatech.edu> at INTERNET
Date:    5/13/99 11:27 PM


You should be able to specify 'localhost' or '127.0.0.1' as the host 
name (or ip). Thus, no code changes will be necessary.
     
You may want to look at SmartRefStream and HTTPSocket - you can send 
objects over, without having to invent a protocol.
     
I'm redesigning PWS to allow a greater degree of pluggability: namely, 
you'll be able to POST various formats, not only form data (regular and 
multipart). For example, you may want to post something in XML or Squeak 
SmartRefStream. Currently, PWS tries to parse it as form data.
     
Could you provide some details about your application? I'm interested in 
creating some fairly generic support for distributed applications, and 
would like to cover several basic and not-so-basic cases.
     
--
Bolot Kerimbaev
College of Computing
Atlanta, GA 30332-0280
     
     
On Thu, 13 May 1999, Jerome Garcia wrote:
     
>      Please forgive my lack of familarity with sockets in the following 
>      question.
>      
>      I would like to write a distributed application in which two or more 
>      images communicate via sockets across a network. However, I would also 
>      like to be able to run all the images of the application and have them 
>      communicate via the same socket code on a machine (NT) which is not 
>      networked. Is this possible? If so, how do I specify the addresses and 
>      ports for the non-networked machine? If it is not possible, how can I 
>      accomplish the equivalent and keep the code differences at a minimum? 
>      
>      Thanks in advance.
>      
>      Jerome
> 
>