Data types and default values

[Robert D Anderson <robander@us.ibm.com>]

Hi,

A couple of reviewers have commented on the data types and default values
in the language specification.

Comments from a couple of reviewers include:
* What do CDATA and NMTOKEN mean?
* What does #IMPLIED mean?
* Why does it say the default is #IMPLIED when my processor always does XYZ
with it?
* Why does it use DTD syntax (#IMPLIED) instead of Schema syntax, or
something more general?

My question is now - what should these values say? Considerations are:
* CDATA, NMTOKEN, etc do have external definitions.
* How can I explain these in words without overloading the attribute tables
with definitions? That is, we do not want each attribute to define NMTOKEN
or CDATA
* I know many look to the default column to see what DITA defines as a
default - but processor defaults are distinct from DTD/Schema enforced
defaults

My preference is:
* For data type, stick to CDATA/NMTOKEN, because these are common XML
terms ... unless somebody can suggest something better. Maybe "Character
data" or "String" for CDATA, but NMTOKEN?
* If needed, we can define those terms at the start of the spec (or
reference definitions elsewhere)
* If there is no DTD/Schema default, list "No default" or leave the column
empty instead of listing #IMPLIED or #REQUIRED
* I think that the "default" column should stick to listing only DTD/Schema
defaults (not processor defaults), otherwise we run in to trouble with
vendors / processors

Other thoughts?

Robert D Anderson
IBM Authoring Tools Development
Chief Architect, DITA Open Toolkit