From: Martin D. <mar...@te...> - 2003-12-23 17:59:16
|
More Java interfaces has been published online. Usual link: http://geoapi.sourceforge.net/javadoc/ There is some issues related to UMLs: * The UML should specifies which CharacterString are localizable. European Union (and other countries like Canada) will require localizable softwares, some *by law*. My suggestion: introduce a new stereotype of UML class, LocalizableCharacterString. UML would need to be updated accordingly. * Is org.opengis.sc.CRS really needed? This interface stands between ReferenceSystem and CoordinateReferenceSystem and break an otherwise quite natural hierarchy. Furthermore, its only "raison d'=EAtre" seems to be related to CompoundCRS contract ("A Compound CRS is a coordinate reference system that combines two or more coordinate reference systems, _none of which can itself be compound_" - emphasis is mine), which I question. I will expose my feeling if there is some response on this topic. * Where are gone CoordinateSystemFactory and the like? Those factories were part of "Coordinate Transformation Services" (document 01-009 by Martin Daily). * Why useful methods from "Coordinate Transformation Services" do not appears in UML (e.g. CoordinateSystem.getDimension())? Generally speaking, should we try to incorporate 01-009 good ideas into GeoAPI or should we stick to UMLs exactly as they are? * Why 'name' and 'ID' attributes have classname prefix in every interfaces (e.g. 'axisName', 'ellipsoidName', 'csName', etc.)? They are the only attributes to follow this strange policy in the package I have seen up to date. It make it harder to write a common tool for discovering arbitrary object's name and ID (for example in an user interface), for example through the 01-009's org.opengis.cs.CS_Info interface. Many of the above issues are flagged in the Javadoc with @revisit tag.=20 Comments would be appreciated. Regards, Martin Desruisseaux. |