Update proposal 2008 08 17

In Early 2007, XDMF underwent a substantial revision. What was the original library we now call XDMF1, and the new version is called XDMF2. The move was enabled by a switch to the libxml2 xml parsing library. Changes to XDMF included:

 1. Replace expat + Xdmf functions with libxml2
 2. Support XPath
 3. XdmfStructure + XdmfTrasform have been merged into XdmfDataItem
 4. DataItem support hierarchical structure
 5. Grids  support hierarchical structure
 6. Added support for Quadratic elements
 7. Added support for mixed topologies
 8. Added XdmfInformation - Application specific data
 9. Added Writing capabilities to Xmdf lib. The state of the DOM is
      accurately reflected in the XML
 10. All XDMF XML Elements now have an XdmfObject representation
 11. Adding DSM
 12. Update from hdf 1.6.2 to hdf 1.6.5

At the same time VTK's (and thus ParaView's) interface to XDMF was partially updated to use the new library. That is, vtkXDMFReader was upgraded to recognize the XDMF2's new syntax and directly produce composite data sets instead of multiple outputs, as well as to understand quadratic elements and mixed topologies and produce the corresponding vtk data structures. Meanwhile, vtkXDMFWriter was only upgraded enough to use XDMF2 in a backwards compatibility mode.

Since that time important new features have been added to XDMF2. Although these features exist in the library, not all of them are useable in the vtkXDMFReader, and even fewer are useable in the vtkXDMFReader. The new features include:

temporal support - Jerry and Biddiscombe (Early 2008)
additional cell types - CSimsoft (Early 2008)
ghost levels - Chris Kees (May 2008)
sql - Jerry (?)
compression - Ian Curlington (May 2008)
parallel "strategies" - Will Dicharry (Jun 2008)
in place row/major column reordering - Dominik Szczerba (Jun 2008)
better hdf version independence - Jerry (Jun 2008)

Additionally interest for the following features exists:

 units annotations
 tabular data (info vis)
 out of core row/major reordering
 SIL interface
 static geometry and topology time varying data optimizations
 wildcard specification

Finally, a full suit of web resources for XDMF would be beneficial to the community at large. The most important is comprehensive regression testing. All fundamental data types should be regularly tested so that we get decent coverage of the libraries capabilities, and so that new users have a set of examples to work from. Additional resources such as a bug tracker and online would doxygen documentation are desireable.

WRITER PLAN OF ACTION

 Produce hierachical data items when given composite data object inputs
 produce full set of data types that reader recognizes
 add support for time
 parallel efficiency