MI SDTS format... (2)

[Malcolm Williamson]

Rich Shepard wrote:
(lines deleted)
>   That's the problem for raster data, too. For example, cells could have
> attributes for soil, depth-to-bedrock, slope, vegetation, aspect,
> erodability, and so on. In my opinion, a GIS cannot do meaningful analyses
> or modeling without a solid, relational database in which to store
> attributes. Mapping, yes; spatial analyses/modeling, no.
> 
(more lines deleted)

Rich,
I think you're approaching GRASS with a little too strong of a
relational database mindset, at least on the raster side of things. The
GRASS raster format already supports multiple "fields" or attributes for
a cell... they're called reclasses. Remember, all r.reclass does is
create a lookup table. It is possible to have thousands of attributes
for the same cells, and still have efficient data recall. I have found
this model to be every bit as functional as the linked database model
used in some competing raster packages, and I would dare to say that
there are quite a few other serious modelers out in GRASS land that
would agree.

Now for vector data, with it's significantly more complex topology,
etc., an rdbms data model makes a lot of sense to me.
-- 
Best regards,
	-Malcolm

Malcolm D. Williamson                       malcolm at cast.uark.edu
Center for Advanced Spatial Technologies      Voice: 501-575-2734
12 Ozark Hall                                   Fax: 501-575-5218
University of Arkansas, Fayetteville AR 72701      
http://www.cast.uark.edu/