[Coin-standards] An argument for public domain modeling infrastructure

[Leonardo B. Lopes]

On Fri, 12 Apr 2002, Irv Lustig wrote:

> But the industry could decide to do something similar.  Let's say a new 
> modeling language called "LEO" (Language for Enterprise Optimization) was 
> designed.  Let's further say that there is a LEO Board, that has 
> representatives from the vendors, industry, and academia.  The Board 
> decides what the language looks like.  The academics create the 
> implementation prototype, and put it out in open source. The vendors take 
> that prototype and have the right to improve it.  They have the option to 
> put back their contributions into open source.  However, if the vendors 
> want to call their language "LEO", it must conform to the standard agreed 
> to by the LEO Board. Why couldn't this process work?

I think there are too many technical, human, and strategic issues that
would prevent an initiative like that from working. In addition, I don't
think a magic bullet modeling environment can be built. Thus the academic
contribution would be limited, so I don't see how academics would be
interested in working on it. What they certainly would be interested in,
as much as practitioners would, is in modeling language integration. So
let me morph your idea just a little and let me know what you think.

The morphed idea is that the modeler would be able to use whichever
modeling environment is more appropriate for the situation, but access
certain objects -- those that can be generalized -- always with a
consistent interface. Lets call the interface "IRV" (Inteface for Real
Value). Through this interface, the modeler might be able, for example, to
fill in parameters, examine variables, relax constraints, solve, obtain
results, etc... 

The modeling language would generate objects which implement this
interface, or would implement it itself. If IRV is not artificially
restricted by IP issues, a modeling environment designer could choose to
implement it or not. IRV would address a lot of the issues you consider
design shortcomings in current AMLs, protect clients' IP, eliminate the
need to port code to a new language, and help (along with an instance
standard) to foster new modeling tools.

Something along those lines was in my proposal, although it doesn't look
like it will make it to the dissertation. But I'd be interested in
pursuing that avenue in the future. Maybe next year.

I think LEO and IRV are taking the discussion in a completely different
direction, though. It is a very interesting direction, but I'd like to try
to move back to the instances for now. After all I do have to get out of
here and get a real job. Besides, both LEO and IRV would benefit from a
more robust instance representation and exchange mechanism.

========================================================================
Leonardo B. Lopes                                       leo at iems.nwu.edu 
Ph.D. Student                                              (847)491-8470
IEMS - Northwestern University              http://www.iems.nwu.edu/~leo