[freeciv-ai] Re: The case for a client AI [was: omniscience problem]

["Per I. Mathisen" <per@xxxxxxxxxxx>]

On Wed, 9 Apr 2003, Mike Kaufman wrote:
> > I agree it shouldn't cheat the rules, as in
> > revolution in one turn, any tax rates whenever and change production at
> > will, but omniscience is another thing altogether.
>
> oh, you're proposing something that doesn't cheat the rules?

I'm merely suggesting we make its existing omniscience consistent. It
places cities based on omniscience, then fails to utilize necessary
squares due to lack of omniscience. This is bad.

> Think of 'not omniscient' as the first but large step toward a client ai.

I am interested in writing a client AI. But this is at least a year away,
and it will be a totally different kind of AI. The lessons learned in
improving the server-AI will come to good use, but much of the actual code
cannot be ported. The two problems omniscience and direct access to server
internals ensure this. An AI that has to solve both lack of knowledge and
network synchronization issues will need to be structured very
differently.

I too was an optimist on behalf of turning server-AI into client AI, until
I realized the magnitude of the changes involved.

Of course, if anyone wants to write a client AI, they are welcome to do
so. I know Raimar has certain long term plans in this area :)

  - Per