Swarm SFI-Hive Pheromone, October 1998

[Marcus G. Daniels]

                     The Swarm SFI-Hive Pheromone
                          Volume 3, Issue 6
                           October 23, 1998


   The Pheromone will stand to inform users of the activities and
goals of the hive and user community.  Contributions are accepted that
announce Swarm-related events or activities in any of the growing
Swarm colonies around the world.  To contribute, send e-mail to
address@hidden  The Pheromone is mailed out if and only if there 
is information to be disseminated.


Table of Contents
-----------------
   I. We are alive and well     
  II. Version 1.3.1
        Current status 
        How to order the CD-ROM
 III. Beyond version 1.3.1
        Version 1.4
        Goals for the coming year
  IV. Swarm Program News
        New program offices 
        Funding update 
        Next Swarm User Group Meeting
   V. Swarm Advisory Board news
  VI. Research update
        AlChemy
        Swarm/SME federation

========================================================================
   I. We are alive and well 

There has been much change in the Swarm Program in the last half year.  
Chris Langton (the founder of the Swarm Program) and Glen Ropella have left 
SFI to start thier own company, The Swarm Corporation. The Swarm Corporation 
seeks to develop new agent based technologies.  Irene Lee left SFI in June to 
take some family time and pursue a full-time position in project management. 

Marcus Daniels and Alex Lancaster remain at SFI as Swarm programmers.  
Alex, also at SFI, splits his time between research applications (see
Research News section below), being the liaison between SFI and external 
researchers on new Swarm projects and core Swarm programming. Roger Burkhart 
remains an active member and chair of the Swarm Advisory Board.

In addition, we have hired a new programmer, Vladimir Jojic, who joined
the Swarm team on October 19th.  Vladimir hails from Belgrade,
Yugoslavia where he completed research in the areas of functional 
programming and natural language processing.  His talents and previous
experiences make him a good match to the needs of the Swarm program.

  II. Swarm version 1.3.1 

Swarm version 1.3.1 was released on October 16th 1998, the new features 
include:

  o A Swarm CD-ROM, formatted for use with Unix and Windows including 
    complete HTML documentation.
  o A simple animated schedule browser. 
  o Much more complete and robust widget snapshot features
  o EZBins are more colorful customizable.  
  o New functions initSwarmApp and initSwarmAppArguments. 
  o Support for char, short, unsigned short, long, and unsigned long
    types in Message and Variable probes.
  o More complete bookkeeping and descriptive statistics for Zone.
  o Various autoconfiguration improvements:
  o Numerous bug fixes
 
To order the CD-ROM please send a check with your full return address and 
mail it to:

    Swarm Program
    Santa Fe Institute
    1399 Hyde Park Road
    Santa Fe NM 87501-8943
    USA

Please make checks out to: "Santa Fe Institute" and include the phrase
"Swarm 1.3.1 CD-ROM" in the memo portion of the check. 

 III. Beyond Swarm version 1.3.1

The immediate goals for version 1.4 include enhancements and test
suites for the activity library and support for handling large and
complex data files.  http://hdf.ncsa.uiuc.edu is one likely substrate
for these features.  Marcus will be working on the data handling, the
core bits of test suites.  Alex intends to work on pedagogical
activity library examples for users (that also serve as tests), and
Vladimir is already intensely engaged in verification and extensions
for the activity library.

Paul Johnson, currently in residence at SFI for two weeks, is
working closely with Alex on a first-draft prototype Swarm User Guide -- 
contributions are welcome, please contact either Alex <address@hidden> 
or Paul <address@hidden>.

In November, Vladimir and Marcus will work on distributed multi-language 
integration features.  We hope to support both CORBA (which will provide 
coarse-grained method callin/callout for most languages) as well as custom, 
high-performance language integration for (at least) Java.  To this end,
we will probably use http://sourceware.cygnus.com/java, since it will
provide link-level object code integration, i.e. performance like calling
Fortran from C.

We are happy enough with Tk portability, that we now put Java language 
integration at a higher priority than Java Abstract Windowing Toolkit GUI 
support (AWT).  However, significant work has been done over the last year 
on cleaning-up the GUI of Swarm, so AWT support, or GTK, etc., will be easier 
when there are evident benefits.

Overall, we intend to make writing Swarm models in Java not only possible, but
efficient and seamless as well.  Now that interfaces to Swarm are uniformly 
`protocol' based, Swarm capabilties are advertised in a conservative, basically
language-neutral format.  We have mechanics in place for reading these per-
library protocol files and generating RPC stubs and header files for arbitrary 
languages.  We do not have a rewrite of Swarm *in* Java as a goal. 

Over the next year:

The top priority for the Swarm is to reduce the amount of programming 
knowledge needed in order to develop a useful simulation.  To facilitate this,
the Swarm team at SFI will be developing interactive activity browsers and 
integrating with popular visual programming environments.  The multi-language
capability will provide the ability to write agent models in languages other
than just Objective C.  In order to support large and complex models in a
scalable and transparent way, the Swarm team will add object distribution
and parallelism features using the existing logical Swarm model.
This will complete the original design of Swarm as a usable and efficient 
fine-grained simulation system.


  IV. Swarm Program News

The Swarm Program recently moved from the gatehouse to the Main Campus at SFI.
Vladimir is sharing office space with Marcus, while Alex is a short 
distance down the hall.  The Swarm programmers have had their contracts 
extended until October 1999.

We would like to thank JWAC and David Lynch for their continued support of the 
Swarm Program.  We are still seeking additional funds to help underwrite the 
continued development of the Swarm Simulation System.

Several community members have asked what is in store for Swarmfest '99.
We would like to begin rotating the venue for Swarmfest and host the meeting 
somewhere other than Santa Fe.  We are considering moving it to UCLA for next 
year since there is a sizeable Swarm contingent there. 
Your suggestions are welcome.  If you would like to volunteer to be a member
of the program committee, please contact Alex Lancaster (address@hidden)
or Irene Lee (address@hidden).


    V. Swarm Advisory Board News

The Swarm Advisory Board has reformulated its role and now serves strictly 
as an advisory board to the Swarm Program at SFI.  Its goals are to assist in 
the management and direction of the Swarm effort at SFI.  Its main focus is on
the continued development of the core Swarm release, but the board will also 
help to plan SwarmFest and other user community events.


   VI. Research News

Amongst the new Swarm research initiatives at SFI is a joint project
with Alex Lancaster and Walter Fontana, Research Professor at SFI
extending previous work done on AlChemy (ALgorithmic CHEMistrY).  The
AlChemy project (http://www.santafe.edu/~walter/AlChemy/alchemy.html)
aims to develop a formalization of the structure/action association
called `chemistry' at a level of abstraction that is useful for
understanding biological organization, its origin and evolution.  

The current project intends to extend the previous work done studying
chemical self-maintenance in AlChemy in a point-dimensional
`well-stirred' reactor case; to a spatially extended two-dimensional
system, using Swarm as a substrate for implementation.

Thomas Maxwell and Ferdinando Villa of the University of Maryland
Institute for Ecological Economics announced that they will be starting
work on the integration of Swarm with SME.  This integration would
allow for spatial parallelization of Swarm models.

The task involves writing a `dynamic space' class for Swarm which would use 
the SME client library to fire up a possibly remote SME simulation driver,
control the SME simulation scheduling with explicit methods invocations, and 
access the SME dataspace.