The Weekly Pheromone VII I9 (try 2)

[gepr]

                         The Weekly Pheromone
                          Volume 2, Issue 9
                            June 19, 1997


   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 on Tuesday (or....
maybe Wednesday or Thursday) if and only if there is information to be
disseminated.

Table of Contents
-----------------

   I.  Future Of Swarm
  II.  Platform Independence
 III.  The Swarm-Dev list
  IV.  Contributions
        
========================================================================

   I.  Future Of Swarm (encl. swarm.org.jpg)

     We now have a pretty firm idea of how we would like Swarm to
continue as an organization.  It has some nice aesthetic qualities to
it because the model of the organization, itself, falls into the Swarm
framework, which any good agent-based system should, eh? [grin] What
I'm about to outline is the *model* of how we think things need to be
to instatiate a distributed software development effort.  Stating it
here by no means implies that we've achieved it, yet.  Here's how we
envision it (I've attached a jpeg of the organization to the end):

1.  There will be a nonprofit organization called the Swarm
Development Group (SDG) that will consist of a board of directors
whose primary responsibility will be to provide strategic direction to
the development of the Swarm software package.  The SDG will be set up
as a legal entity separate from the SFI or any other institution.  It
will have no assets except for, possibly, the copyright for the Swarm
software.  It won't have any employees and it won't really have a
permanent location.  The board for the SDG will consist of Roger,
Chris, me, Barry, Sven, persons from each major development site, and
a smattering of others from the user community.  We're going to set
the board up so that it will have something like 11 or 13 seats.  We
plan on making Roger the Chairperson, since he holds the grand vision
for a public domain, distributed software development environment.
The board will meet (physically or virtually) as often as necessary to
handle any issues that may arise.

    The idea is that the SDG will provide the locus or "hub" for the
user community.  It will be the body that ensures the quality and
integrity of the software package by making decisions about what
long-term directions the package should take and what processes are
used to develop the package.  This includes decisions like what a
parallelized Swarm needs to be able to do, what the software process
should be for integrating new code, and what domains Swarm should
actively pursue.  It doesn't, however, include the day-to-day
administration of how one achieves parallelism, implementation of a
software process, or writing code to implement domain specific
libraries.  These latter functions will be done by the various
development sites.

2.  A primary development site, called the Swarm Development Program
(SDP), will be seeded at *some* institution to carry the ball from
where we will be at the end of September '97.  This group of people
will consist, minimally, of 2.5 people: a Program Administrator (half
time), a Kernel developer (full time), and a Library developer (full
time).  The SDP will be an effector of the SDG, by which I mean that
the SDP will implement the software process, do work on the kernel and
libraries that is required for parallelization, etc.  At first it will
probably be the only effector.  But, as time goes by, hopefully, other
effectors will arise.  Some sites that are already close to being
effectors are UMich, UCLA, and some individuals who contribute
significantly already.

    The SDP might be hosted by SFI, or it might be hosted elsewhere.
Talks are in progress to find a good location.  The SDP will be in
place by the end of September.  Regardless of where it's located, we
already have two corporations that have contracted for work to be done
on Swarm and we expect there to be more in the near future that will
help cover the costs of the core program.  If anyone would like to
help find funding or has any ideas of relevant sources of funding,
please contact one of us.


                        ----------------------

  II.  Platform Independence

    Intel Corporation has funded an effort to make Swarm "platform
independent."  The motivation behind such an effort is to allow
Windows NT users the ability to run Swarm without switching to Linux
for their operating system.  To achieve this, we decided that most of
the difficulties in getting Swarm to run on a Windows OS lay in the
GUI for Swarm, which consists of X, Tcl/Tk, and BLT.  The minimal
solution to getting Swarm to run on Windows NT would be to use the X
and Tcl/Tk tools provided for that operating system.  But, that would
largely be a one-off hack and would be very brittle.  So, we decided
that replacing the Swarm GUI with one written using Java would be a
better solution.  It would address the Windows NT port and might open
the door to other ports like to a Macintosh or OS/2 system.  Another
major motivation for choosing the JavaGUI route is that doing it this
way will allow us to keep only one revision branch of Swarm.  If we
created a one-off branch of Swarm for NT, which is what would happen if
we stuck with Tcl/Tk, then any changes made to the UNIX version would
then have to be mirrored in the NT version.  This type of thing can
lead to an administrative overhead that is so high as to displace
productive work that could be done on the core package.

    This effort then consists of separating the user interface pieces
of Swarm from the kernel and core libraries, which is something we
need to do for cleanliness sake anyway.  Once that separation has been
done, we can begin replacing all the Tk GUI pieces with Java pieces.
The one caveat is that Swarm now relies on a separately compiled
library called libtclobjc, which provides the glue between Objective C
and Tcl/Tk.  An analogous interface between Java and Objective C has
to be written for this to work.  So far this doesn't seem to be a
problem, though.  Java's native method interface allows Objective C
calls on Java object methods and vice versa.

    We've contracted with a Java expert and all-around
computer/operating systems guru to write the Java GUI objects and help
us link them in with the core libraries (in Objective C).  His name is
Doug Orr and he is accessible via email as <address@hidden> and he's
on the swarm-support mailing list.

    One final note: The kernel for Swarm will still be written in
Objective-C and we will still be using the GNU Gcc compiler, which we
will distribute with Swarm.  But, it will now be dependent on a Java
virtual machine and application environment instead of the X, Tcl/Tk,
and BLT dependencies.  While we are certain this will lead to OS
portability in the long run, it will still depend on the robustness
and efficiency of the Java environment used.  And it won't allow
remote execution of Swarms until we have integrated a remote execution
system into Swarm, itself.  But, it is certainly a step in the right
direction and we are very optimistic about it.

                        ----------------------

 III.  The Swarm-Dev list

     I'd like to remind everyone that the Swarm-Dev email list is
a dead list.  Apparently, there are still some old FAQ's or somesuch
for newsgroups out there that instruct possible Swarm users to join
this email list.  But, it has been replaced by Swarm-Support,
Swarm-GIS, and Swarm-Modelling.  DO NOT subscribe to this list and
expect to see any traffic.

     Just as a refresher:  you can subscribe to any of our four lists
(swarm-support, swarm-gis, swarm-modelling, or swarm-announce) by 
sending a message to:

   address@hidden

with the following in the body (filled out appropriately):

   subscribe swarm-xyz address@hidden

                        ----------------------

  IV.  Contributions

     I'd like to reiterate that any contribution of ideas for future
development, thoughts on the structure of the SDG/SDP, funding
sources, or anything else are absolutely welcome.  We're at a bit of
an inflection point in the event trace of Swarm and now is the time
when the users can and will have a great impact on the future.  As
Warren Zevon might say, "Send lawyers, guns, and money!"



                              =========

The Swarm Pheromone is created by the members of the Swarm project at
the Santa Fe Institute.  For more information on the Swarm Project,
see http://www.santafe.edu/projects/swarm.  Comments, corrections, and
contributions should be sent to address@hidden