DFDL Notes:

Beckerle - DFDL Summary presentation

could go to XML schema group to solve validating typedescriptor-in-annotat ion to XML type validation issue

How much does 'pretty' matter?:

Why not RDF for annotations? - versus XSD annotation? Need a FAQ item (action:Mike) and further discussion

Why is verbosity a problem (readability or performance) - readability

Doesn't tooling solve the verbostiy problem? Yes, but - readability outside tools is still important

Is readability an issue with numeric data (the XML is already hard to read if there are big arrays of numbers in it) - maybe (schema are still somewhat readable/textual)


Jim - presented PNNL/NCSA prototype, DAI-X service usign DFDL

Session 2



Annotations

-IBM IIS proposal
Issue relating to layers - params associated with types versus readers (e.g. bytes-string-int, some params apply to byte-string, some to string-int - hard to capture in a single 'type')

Jim likes the mixin model and thinks it helps with multilayers, etc. - mixins represent transforms/readers, yellow boxes define a (re)usable stack of transforms (could be one).

Action (Mike) - get opinion from CAM model backers of new model

Note that mixin model helps separate 'implementation' concerns that otherwise get embedded on the model (is there an int type corresponding to a class that has text/binary and byteorder flags, or textint and binint type/classes? with the mixin model, both could be supported by users as they like and DFDL can endorse a best practice with our choice of built-in types)



Layering

review of multilayer model discussed previously by phone/email...

generally fits with mixin model - need new base class that does not have byte-order properties, just source indirection, etc. (Mike has details)

redo example - question how would this example get wrapped up into a new 'yellow box' in the mixin model.

Questions - Neil is looking at DFDL use cases in NextGrid, etc. related to OGSA-DAI, ...