REPLY SD Impact on National Government Policies (SD6188)

[System Dynamics Mailing List]

Posted by  Bill Harris <bill_harris at facilitatedsystems.com>

Jaideep,

It's good to see you here again, even if you are from the dark side
now.   :-) 

For the last 10 months, I've been using a free simulator professionally
and for fun.  I've made a well-received presentation regarding its use
at a conference.  I'm noticing a number of characteristics, among them:

  It works on any platform that has gcc and the GNU Scientific Library:
  GNU/Linux, OS X, Windows, and probably others.

  Models are text, not binary, files (one of the complaints occasionally
  raised here).

  The process of modeling may be faster, thanks to the process it
  encourages.  The process seems slower at the start, but I think that's
  because it's forcing me to think hard about the modeling from the very
  beginning (with a GUI, I can always start laying down a stock or two
  before I really start thinking, for I know I'll always need a couple
  of stocks).

  Simulation is generally much faster, for all models are compiled.
  Then again, I run the simulator _far_ less often than I have with
  other systems.

  Because the simulator is GPL'd, you can't ethically or legally
  distribute compiled models (well, perhaps you can under certain
  conditions, but I think that means you'd need to distribute all the
  sources and build instructions for all the libraries that get compiled
  into the code, and you'd need compatible licenses for all that code; I
  don't think that's feasible legally, at least under Windows).  If you
  want to send a client a model for them to run, you need to help them
  install the simulator and compile the model themselves.

  The simulator is open source, so I can see exactly how a simulation is
  implemented and modify it, if I wish.  Do you know what interpolation
  algorithm is used for tabular nonlinearities in the simulator you
  currently use?  Would you like to use another interpolation algorithm?
  (I admit that this usually isn't a big concern.)

  There is (almost) no more concern with TIMESTEP or DT!  (If there is,
  you'll know it, and then you can address it.)

  It's a bit more of a pain to implement discrete parts of models, and
  tabular nonlinearities do look a bit arcane.
  
  It's got arrays and powerful statistical tools.

  To enjoy this simulator, you probably need to be a person who enjoys
  *nux-style work: text editors, command line interfaces, and the like.

  The code is maintained by a group that uses it for serious
  (health-related) work, so they engage in careful quality control.
  While it's been around since at least 1992, as best as I can tell,
  I've not found anyone else in the SD community using it.

And, most importantly for me, I sense this simulator is helping me think
better.  Compared to when I used an integrated, GUI system, I think I'm
designing models more carefully, I'm designing better experiments, I'm
finding new, creative, and useful ways to analyze outputs, and it's
potentially easier to prepare communicative reports.

If you're curious, I've blogged about it a couple of times in the past.
See
http://www.facilitatedsystems.com/weblog/2006/11/system-dynamics-with-mcsim.html
and the links off that page.


>> 3. Some people say good SD is hard, but then programming and designing

Have you seen Druid (http://druid.sourceforge.net/)?


>> 4. The returns from process improvements and usage of even simple

"The Joy of Thinking Small" made it into the top 10 list of my blog
(http://facilitatedsystems.com/weblog/2006/12/top-postings-of-2006.html). 

Bill
- -- 
Bill Harris                   
Facilitated Systems                              Everett, WA 98208 USA
Posted by  Bill Harris <bill_harris at facilitatedsystems.com>
posting date  Sat, 13 Jan 2007 22:03:34 -0800