The relationship between Geotools and J2SE ImageIO requires refinement before too much more code is written. Specifically, the automated discovery and selection of a data source's dimensionality is not necessarily a geographic function. The way to get the most reuse out of code is to provide the most functionality at the most general level possible. In the case of ImageIO and Geotools, this would mean adding descriptive and control capabilities to ImageIO and utilizing this functionality from within Geotools. In this manner, multidimensional plugins written for use with Geotools would have a much wider audience, and multidimensional data sources unrelated to Geotools might plug directly in.
Topics for discussion
- What data discovery already exists at the J2SE ImageIO level? Can we just steal something?
- Should we create a 2D GCE wrapper around ImageIO before adding data discovery?
- Create a standard for IIOMetadata objects returned by GCE plugins, which permits the creation of CRS objects?
- If we need to design our own addition to the ImageIO API, what issues arise?