Problem with r.cost?

[W. Fredrick Limp]

Recently Jianping Xu responded to a question from Rick Thompson about
changes in r.cost from 4.0 to 4.1.  Thanks for this info.  I may have
missed something but it does not appear that the 4.1 documentation
discusses this.  Xu discussion is useful but I wonder if we could
have a more detailed discussion from someone (the reviser?) as
to the specifics... r.cost is a very valuable module and I, for
one, would like to be sure that I know what it is doing.

Some specifics listed below...

>Rick writes 
> > 
> > Hello Grassers,
> >   Does anyone know why r.cost is behaving so much differently in 4.1 than
> > in 4.0. The manual entry in 4.0 makes sense, but after using 4.1 r.cost,
> 
>Xu responds.. 
> Several reasons:
> 	1. 4.0 r.cost figures the cumulative_cost for a to_cell by simply adding
> 	   its cost to the from_cell's; while 4.1 r.cost does it
> 	   by adding the average_cost of the 'from' and 'to' cells to the 
> 	   from_cell's cumulative_cost.
 -----  I follow this...

> 	2. *Absolute* cell resolutions are taken into accout by 4.0 r.cost 
> 	   ('normalized' to 100m), not by 4.1 r.cost. 
 ------- This is not clear. What does "taken into account" really imply

> 	3. 4.1 r.cost takes more 'neighboring' cells into accout as your 
> 	   request, while 4.0 r.cost doesn't. And ...
  -----  I assume this refers to the knights move or perhaps not ????
> 
> Jianping Xu 
> jianp at ocean.rutgers.edu
> 

-- 
W. Fredrick Limp,   Director                     FAX: (501) 575-3846
CAST, Center for Advanced Spatial Technologies   TEL: (501) 575-6159     
12 Ozark Hall, University of Arkansas         
Fayetteville AR 72701                             fred at cast.uark.edu