Re: Sharing Models, was: Intestinal Fortitude

From: Eugene.Leitl@lrz.uni-muenchen.de
Date: Sat Mar 17 2001 - 07:54:23 MST


Spike Jones wrote:
 
> 1) What open source language do we use?

I'm not a rabid Pythonista (rather a Schemer/Forth wannabee), but
I think the sum of power/expressivity, readability/maintainablity,
availability and user community is better for Python than any other
languages around. It's not for high-performance numerics, but that's
what extending Python is for. SPaSM and MMTK would come to
mind here, that's as deep number crunch as it gets.

I wouldn't call Java OpenSource, but if you want to go Java,
consider Jython.

> 2) How is this software package to be accessed?

Sourceforge, of course.

> 3) What architecture?

Which machines should it run on? Why, on anything with an ANSI C
compiler. The software architecture? OO, of course. The embedded
webserver/Apache mod_XY idea is certainly nice. I would try creating
independant components which talk to each other through sockets,
probably using some standartized interface, if that's not too bloated.

> 4) Are there existing open source components that
> we could incorporate?

I don't know what we're needing. Matrix algebra? Infinite precision?
Vizualisation/GUI? Python would seem to work.

> 5) Are we looking a simulation of some sort?

A what if approach certainly seems useful, you can mount higher-order
optimizers on top of that.

> 6) What kind of user interface? Graphic? etc.

Sure. The server-side approach will limit interactivity, though.
You could enhance that by going client-side, but this will break
portability.



This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:59:41 MDT