• Simone gave a history of his involvement with Geotools and the development for the NATO project. This should be placed on the Coverage plans page.
  • OSSIM is interested in working with us. This perhaps makes the transition to ISO19123 more urgent.
  • Martin estimates ISO19123 interfaces approx 40% complete. Needs time to finish.
  • ISO19123 seems generally acceptable for an internal representation of nD datasets. However, it does not address storage.
  • GCE is viewed as generally inadequate for persistence of nD datasets. We'll have to get something working on a branch, then propose it to OGC.
  • JAI discussion, Simone notes phenominal speedup if reading tiled tiffs. Improving grid coverage with smarter JAI aligns with his interest.
  • Martin targets October as completion for ISO 19123 interfaces.
  • New GCE must have following characteristics:
    • Persist to:
      • Stream (including File)
      • Database
    • Implement in a stepwise fashion, streams first, Db second.
    • Tiling should be transparent.
  • IIOMetadataFormat: everyone likes it. Simone already made a GeoTIFFWriter to use it. He's adapting Grib to use it. We just need a standard now.
  • For communication with IIO plugins, we'd like:
    • A common set of ImageReadParams for communication with the plugin. Requirements for ImageReadParams include:
      • time
      • z level
      • Any other way to specify a slice in a N-dimension coverage.
    • A common definition of the IIOMetadataFormat for CRS. Requirements include:
      • CRS encoding
      • Envelope encoding
      • "sample to geophysics" formulas encoding
      • Must be compliant with OGC specifications. Candidates are OGC's 05-027r1 ("Image CRS using GML") and "GML in JPEG 2000" specifications.
    • Helper code to construct CRS from the IIOMetadata.
    • Common classes (not specific to any partocular plugin) in package.
  • J2SE allows for IIOMetadata to be associated with either a stream (file) or a dataset. Simone suggests:
    • Stream metadata = whole dataset (2d, 3d, 4d, 5d)
    • Image metadata = 2d slice.
  • On "slice extraction" topic, we have to clarify where the GridCoverageExchange work stops and GridCoverageProcessor work begins. Martin suggestion is that GridCoverageExchange extracts slices if the requested slices match the underlying data layout. For more complex slices extraction, it would be GridCoverageProcessor work. This doesn't means that a 3D coverage needs to hold fully in memory in order to performs a slice extraction, since we should get deferred execution to work as much as possible.

Action Items

  • Simone, Martin should edit Coverage plans page with their own list of requirements, personnel, etc. Maintain this as time goes on.
  • Simone: Make a wiki page of JAI tips.
  • Alex Petkov : Examine Image CRS using GML to determine applicability.


Full transcript available here .

Posted by cholmes at Jul 29, 2005 05:26