Dear Russ, I agree we want to be able to provide string-valued axes which give "region" as a coordinate, for example. However, if it requires definition of a polygon to make this region explicit, I don't think it is appropriate to handle it through the same mechanisms as deal with the coordinates and boundaries of ordinary gridboxes. That seems to be stretching the convention too far. For the purposes of climate data, especially GCM data, we probably do not need to do this anyway. The regions we use are generally rectangular ones, or else can be defined by supplying a separate mask array (such as a land-sea mask). > to list a set of > variables whose values uniquely determine an index for that dimension. > This is analogous to a multi-field key in a relational database > relation. There may be several candidate sets of fields that might > serve as a key for a relation, but only one set of fields is declared to > be *the key* for the relation. Yes, I think this is an issue, which I was trying to write about in my last mail to Stephen and you. At the moment, I am tending to think * The normal one-dimensional axes themselves can always be used as a valid set of coordinates, even if they are just ordinal indices. * The choice of which possible set of associated multidimensional physical coordinates to use (instead of the normal one-dimensional axes) should be made by the application, rather than be stated in the convention. The convention can provide means by which possible axes can be identified, but the particular set required for a certain purpose may depend on that purpose. Best wishes, Jonathan