[Freeciv-Dev] Re: Goto into the unknown (PR#759)

[Christian Knoke <ChrisK@xxxxxxxx>]

Am Donnerstag,  3. Mai 2001 10:57 schrieben Sie:
> Thue wrote:
>
> <snip bizarre goto routes when travelling into the unknown around the
> seashore>
>
> > Why this happens:

Took me some time to understand that :-)

> > Note that if you have an ocean tile next to a land tile you see the
> > cost. To be able to draw this the client knows the terrain of tiles
> > on the edges of the unknown.

Thats a political question. If you're on the sea, you can see
the coast, if you're on the land, you can't, IMO. Drawing
(I checked trident, civ2, and civ2gfx) also only shows the coast
of ocean tiles in some cases, the margin of land tiles remains
in the dark. The client doesn't need to know the type of "unknown"
tiles adjacent to *land* tiles for drawing, does it?

> >
> > The goto then uses this knowledge, as a user also can.

No, the user cannot, in case of land tiles, s.a.

So there's arising an informational problem, when we go to
AI clients, cheating will be possible.

> > What you are seeing is a case where the drawing to not use all the
> > extra knowledge it has.
> > I suppose it could be hacked to only send land tiles next to
> > unknown ocean tiles...

May be.

Jason Short wrote:
>
> A much better solution, IMO, would be to not send those tiles at all.

I would prefer this, because it's clean.

> The client could then be hacked to show the "darkness of the unknown"
> encroaching a little bit more so that it covers up what is currently
> the seashore (thus solving the drawing problem).  I'm not sure that
> this is necessary or even preferable, but I think it's better than
> what you suggest.

Another possiblity for ocean tiles would be, only send a flag
"land / no land at margin", not the exact terrain type. Is this
sufficient for drawing?

>
> If this isn't done, though, then at a minimum the seashore should be
> shown.  If the client knows it's there and can take advantage of it
> to determine the goto route, then the player should also.  While this
> may be inferior (the client/player still knows information they
> "shouldn't"), at least it wouldn't give bizarre, seemingly
> inexplicable behavior like the goto route demonstrated here.

I don't worry about the goto function any more, but about
client cheating. Goto does a good job here.

>
> jason short

Christian

-- 
* Christian Knoke                           +49 4852 92248 *
* D-25541 Brunsbuettel                  Wurtleutetweute 49 *
* * * * * * * * *  Ceterum censeo Microsoft esse dividendum.