sci.op-research

On Feb 23, 11:01 pm, A.L. wrote:
> On Sat, 23 Feb 2008 06:25:38 -0800 (PST), Min
> wrote:
>
>
>
> >On Feb 22, 7:55 pm, A.L. wrote:
> >> On Fri, 22 Feb 2008 12:23:15 -0800 (PST), SteveM
> >> wrote:
>
> >> >If this model is for production purposes, i.e., regular operational
> >> >scheduling then you'd probably want to talk to your IT guys to see if
> >> >they could develop it, with you assisting.
>
> >> And this is the problem.
>
> >> IT people will ask: "Where is the API?". OR specialist will ask:
> >> "Where is WHAT?"... After short discussion, IT will say:" No, we don't
> >> know what is OPL, MPL, GAMS an AMPL. We program in Java". OR wilt say
> >> (proudly, even extremely proudly): "Java what?.. I AM OR SPECIALIST! I
> >> PROGRAM IN AMPL!!!"
>
> >> And this will be end of the story :)
>
> >> A.L.
>
> >> P.S. The above is real story...
>
> >A.L.
> >LOL...
> >Maybe I should give them my codes as open source. Go and make whatever
> >you want. API? Look up in the CPLEX manual;)
>
> >Min
>
> No. IT will NOT do your job. And your job is to give them API.. If you
> give them your code and say "do whatever you want", they will put this
> stuff in trash.
>
> CPLEX API is NOT your API. Your API is API to your specific model. And
> this can be quite complex: it must provide all data, control model
> execution and retrieve results.
>
> A.L.

Version 5 (that I am using) has documentation that describes several
ways: C++, .net, Java. Sorry but I don't understand them as well :-).
However, I guess depending on how 'robust' your API solution needs. I
used OPLRun once, which 'hides' the OPL from the user and with MS
Access as the front end -- easier to implement for a non-so-IT person
like me, but more suitable for a standalone type of application. In
the documentation, it was meant to be for pre-deployment check though.