Agustin Lobo lobo at Jasper.Stanford.EDU
Fri Apr 16 08:59:48 EDT 1993

Good that somebody finally takes this issue seriously! Here you have my 
answers ("Lo bueno, si breve, dos veces bueno." F. de Quevedo, 1575).

1.  What are the requirements for adequate implementation of floating point?
	- single vs. double precision: DOUBLE
	- should measurement precision metadata be attached to a raster, to
	truncate display values, and to be propagated to rasters derived from
	the original data set:IDEALLY
	- should an integer scaling be stored concurrently to benefit rapid
	display, or do the maintenance overhead and storage costs outweigh
	the value:NO

2.  Is there an existing format standard (e.g. HDF) which should be adopted?
	What are it's advantages, disadvantages?:HDF would be great for portability (i.e. SpyGlass).

3.  If floating-point support were to be implemented incrementally (utility
	by utility), which utilities should receive first priority for 
	floating-point support?: r.mapcalc

4.  Are there potential problems which may be anticipated to arise in the
	development of floating-point operation library functions which must
	remain compatible with the existing integer-based routines?

5.  Null data bit maps seem a logical solution to provide distinction from
	true zeroes.  Are there other techniques which should be considered?:
we could adopt NA and INF. The other functions should provide the optio of
just ignoring these values.


More information about the grass-dev mailing list