[Freeciv-Dev] Re: normalize_map_pos and invalid map positions

[Jason Short <jdorje@xxxxxxxxxxxxxxxxxxxxx>]

Ross W. Wetmore wrote:
> At 01:25 PM 02/08/09 -0500, Mike Kaufman wrote:
>>example: a unit at (0,0) is ordered to move "northwest" by:
>>a) the server (by proxy of the ai)
>>b) a remote client.
>>
>>What should ideally happen in a general topological framework?

> > Jason, Ross, could we have a consensus on this particular point, devoid of
> > any other considerations? If other considerations, what do they need to be?
>
> 1) Normalization is a "soft" condition that can be defined locally
>    within a module. Different modules can have different concepts of
>    normalization. Normalize_map_pos() should be used at portal points
>    between such logical regimes to trap errors and do any conversions.
Usually this is correct, but it does not generalize completely.  If we 
say (-1,-1) is a special value, then that value must be preserved 
between modules.  The easy solution is that (-1,-1) should not be chosen 
as a normal value for any module.  The "correct" solution IMO would be 
to add a flag to coordinates passed in this way to represent these 
special values.  This does generalize well, and will avoid having 
(-1,-1) mean one special case while (-2,-2) means something else.
> 2) (-1,-1) is a "special" value in the code. It *cannot* be assumed to 
>    be a coordinate except in context. Changing this is both hard and
>    probably not worthwhile given 1). Finding a replacement is equally
>    dubious.
Mostly correct.  (-1,-1) is not just a special value in the code, but in 
the network interface as well right now.
Note, though, that using capabilities it is easy to change the internal 
representation while preserving backwards compatability.  Doing so would 
require a fair amount of code change, but is probably the easier 
solution in the long run - making fixes like the vnotify_conn_ex one all 
over the code is not a good long-term solution.
jason