[Freeciv-Dev] Re: (PR#4242) Clean up the goto route network protocol and route execution

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

On Fri, 16 May 2003, Raimar Falke wrote:
> > Let's say a piece of road in your path, which covered by FoW, has been
> > pillaged. This means you just lost 3, not 1, MP. When you reach the first
> > series of dangerous tiles, you (the client) thought you could pass, but in
> > reality you can't. So you should have stopped.
>
> In this case the server should IMHO stop after the step onto this tile
> because it sees that the moves_left don't match. It should stop and
> put the unit to idle. Then the client (human or agent) has to produce
> new commands.

Ok, this sounds like a plan. But do give a proper feedback to the user,
like "Goto stopped due to changes in the terrain."

> The goto execution should also be aborted if an enemy stands in the
> path and/or is visible. But there needs to be a possibility to say
> "attack with this move" which will stop if there is no enemy at the
> position.

Yes as a single variable, not per tile. If zero, don't attack if we
encounter a unit at our destination.  Encountering a non-allied unit in
our path before our destination should always be a mistake, and the goto
should abort.

> This gets complex and is one of the reasons why goto
> execution should be done at the client side. At least for the agents.

If you have the above, I don't see why. Even for agents, you can with
relative safety send goto execution to server _as long as you don't reduce
the amount of FoW_.

> The proposed patch is good enough for the human player. So I'm still
> for applying this patch and not expand it.

Agreed.

> If this one is in two more
> patches will come: one to convert connect and go-to-city to the client
> and one for air units. The latter is more Greg's area.

Sounds good.

  - Per