2004-06-07 GGF11 DFDL-WG Session 1
Status Review

The slide presentation was not interrupted by many questions. When we
reached the end a lively discussion occurred.

Topics included:

* As expected many of the questions dealt with the data formats that
  aren't discussed in the primer. E.g., stored lengths were raised by
  someone familiar with a medical image standard format that uses lots
  of stored length representations. Null values were raised, i.e., can
  DFDL handle them and roughly what would it be like.

* Some exploration around the limits of expressive power
  occurred. That is, what can't be described in DFDL. Mike Beckerle
  made the point that it might not be possible to describe XML data
  using DFDL because of the complexity of the namespace stuff in the
  parsing of XML. That is, such a system amounts to reinventing an XML
  parser in DFDL, and there would be no real point in that. However,
  if it's not even possible to do it we want to know why, and we may
  or may not be happy with that limitation.

* The point was made that we seem to be trying to put everything onto
  the type definitions of XSD, and perhaps we don't need to or
  shouldn't push everything onto the type definitions.

* Layered/Nested representations were raised. That is, how to handle
  when there is a structured type (a group), but the representation is
  somewhat different from it. Not alot different, but somewhat
  different. E.g., a group containing 2 int fields logically, but
  their representation is as shorts or bytes.

* What are the difficulties with XML/XSD? Tried to defer this to
  wednesday, but some people wanted to know what the discussion would
  be about, so we pulled out a slide for wednesday listing the topics.

* Jim Meyers made the point that DFDL will be very popular because it
  will free people from having to use more complex API-based data
  representations in order to share data. Instead they can do what is
  easiest and most expedient for them in the short term, knowing that
  the data they're creating can be described in DFDL for subsequent
  users longer term.

* Guy Rixon asked about an API for accessing DFDL described data. We
  discussed that XML-ish APIs would of course be possible, then we
  discussed something more like the SDO specification which has
  recently come out from IBM/BEA.