Mike
Beckerle
Martin
Westhead
Geoff
Judd
Bob
McGrath
This
was a very useful discussion. We started by formulating the outline of the spec
document and then reviewed some specific aspects of the new format document.
Spec document outline:
Introduction section
·
glossary
·
What is DFLD?
·
Model
Language basics
Detailed semantics
Arrays
Built-in specifications
Extensions
Extensibility – constructs like <dfdl:sequence> need to
have their list of allowed properties extensible
Format/defineFormat – is the dfdl:format tag correctly named.
Martin assumed a catch-all dfdl annotation tag. Mike assumed that dfdl:format
was *just* format and that other
things (variable assignments, conversion definitions etc.) would go in
different tags. More examples and investigation are required to see how this
works in practice.
defineFormat
is probably too specific right now, this is an important reuse mechanism and it
should be available to apply to all top level dfdl tags. One solution would be
to have a more generic “define” tag that can contain “dfdl:format”
tags along with others.
Properties and variables – we expect that properties and variables have
the same scoping, assignment and reference semantics. Properties are only
different in the sense that they represent parameters to conversions.
Linked properties – we discussed the need for linked properties, it seems
like extra stuff that might not be required and it was useful to see if there
were different ways to achieve the same goal. From the discussion it did not
seem an unreasonable thing to include.
Applies default – we reviewed the issues around the “applies”
default (toScope or hereOnly) the current default is “hereOnly”
different examples suggest different defaults are more appropriate. We decided
to leave the decision where it was but a bigger selection of examples will
probably help inform this decision.