[Freeciv-Dev] Re: (PR#5438) Starting position tidbit

["Gregory Berkolaiko" <Gregory.Berkolaiko@xxxxxxxxxxxx>]

On Tue, 26 Aug 2003, Karen Yeats wrote:

> Hello.  Back in #2503 the biggest part of the starting position problem
> was fixed, and hopefully now games failing to find sufficient starting
> positions are sufficiently rare that it isn't even a problem for most.  
> 
> However one situation remains in which the game can often fail to find
> starting positions.  That is the situation when the terrain is bad
> almost everywhere (this can be a fun challenge to play).

I dislike the patch for two reasons:

1 (serious)  is_good_starter presumes that SPECIALs have certain effect, 
like SPECIAL_1 in desert is always oasis with good food.  This can be 
changed by rulesets!  And all other code I know can deal with it flexibly.  
I suggest you base your mark on food & production values instead.
Have a look at Per's new settlers patch somewhere on Freeciv-AI.

2 If you start on 256x256 map with plenty of plains and space, the 
algorithm can still put your settler on a forest.  The check that to 
prevent such things is currently
+  /* don't start on bad land until we're getting desparate */
+  if ((map.xsize + map.ysize) / 5 < dist * is_good_starter(x, y)) {
+    return TRUE;
+  }
I would rather suggest a check against the amount by which dist was 
decreased in create_start_positions.  If it dropped 30% from it's original 
values, we can consider forests, 50% -- we will take good hills etc.  
Something along the lines anyway.

G.