Re: Shared game-data

On 30 Apr 2006 20:15:46 -0700, Marshall Spight wrote:

> I only want to add one small point. As you say, the current state
> is that the programmer has only a single language with which to
> express both the logical semantics and the physical implementation
> of his program, in a conjoined way. The desired end state is that
> the programmer specifies his program's logical semantics and
> software provides a corresponding physical implementation.
> There is a state along the way that is worth mentioning.

Is it use cases, requirements, specifications?

Anyway, you can have as many languages as you want, if they are translated automatically into each other. But if you want to leave that burden on programmers, then in the end, you will have nothing but poor quality. This is why the "nGL" idea does not work. If n+1GL does not fully replace nGL, then better n, than both n and n+1.

> It would be an improvement over the current state of affairs
> if the programmer had a language with which to express
> the logical semantics of his program, and a separate
> language with which to express implementation. Thus
> different implementations could by tried out, manually,
> and measured, without affecting the results of the
> program.

Why not to do it in one language?

> I don't know of anything that does this now, but the OOP
> world is slowly groveling its way in that direction.

[I hate UML (:-))]

> For example,
> it's starting to become a common idea in Java programming
> that one uses interface types at module boundaries, and
> only uses implementation types inside classes. For example,
> a method might be declared to return a java.util.List, and
> only the internals of the class know whether it's an ArrayList
> or a LinkedList. Quite primitive compared to what could be,
> but still a step in the right direction.

Yes, but this is one language, and it should stay so, IMO.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de