Changes between Version 2 and Version 3 of GridTopologyIssues

Timestamp:

Apr 18, 2013, 9:01:24 AM (12 years ago)

Author:

Peter Baumann

Comment:

—

GridTopologyIssues

@@ -26 +26 @@
 The result of a WCS Core subsetting may not be a grid (ie, array) any more.
 
+Observations:
+ * we believe that the origin of a coverage does not need to be inside array (as per GML 3.3)
+ * trimming (and slicing?) generate grids where some points need to be existent for reasons of toppology, but do not have any meaning (nor in value, nor do they have a a valid geo position)
+ * !GetCoverage establishes an intersection of domain D and bounding Box B. Intersection must maintain grid property, which is not naturally fulfilled. 
+ * offset vectors do not pose a problem for any coverage type (regular: offsets are the same everywhere, so position independent; irregular axes and GCPs: offsets are absolute, ie, depending on origin only and not on neighbor offset)
+
 '''Alternatives:'''
- * Alt 1:
+ * Alt 1: reconstruct grid by omitting elements of the intersection 
+ * Alt 2: reconstruct grid by introducing dummy nodes (like black pixels in reprojected images)
+ * Alt 3: reconstruct grid by including adjacent nodes outside B
 
 '''Discussion:'''
- * Pro 1:
- * Con 1:
+ * Con 1: loses data
+ * Con 2: dummy nodes also have no valid coordinates (offset vectors or GCPs!), not only invalid data
+ * Con 3: includes data "not wanted"
 
 '''Decision:'''