John Cowan wrote:

>  >
>  > (Y) This is the space where high fidelity roundtripping apps
>  > live. Apps that care about the difference between:
>  >         <name first = "Sean" last = "Mc Grath"/>
>  > and:
>  >         <name
>  >                 last = 'Mc Grath'
>  >                 first = 'Sean'></name>
> 
>  I think it's pretty arbitrary of you to call the Y level "high
>  fidelity", just as it's arbitrary for me to call the b level
>  "the Infoset".  But we both agree that a line must be drawn somewhere.

What we mean by "high fidelity" is simply this: the *parsed* model,  what is
also called a grove, has the capability of saving the document *exactly*,
character for character, from what it was parsed.

>  ...
>  We used to have that explicitly, with a "minimal infoset", a "maximal
>  infoset" and every possible infoset in between.  It was a mess to
>  describe.  ...

But it is not hard to describe, that is my point. The XML 1.0 production
rules create a parse tree which exactly describes the source document, down
to the byte. This is what I would call the XML property set, and from this
one can subset to one's desire. Several issues remain, 1) namespaces --
which require additional constraints as well as information items, 2) base
URI and other desired contextual information not obtainable from the source
character stream.

> 
>  1) one standard infoset
>  2) conformance means documentation of what you provide and what you don't
>  3) any app can add its own info items and properties
>  (right down to the a level if you want)
>  4) the "core conformance" infoset is one possible subset
> 

Doesn't it make the most sense to subset from the full description, rather
than both add to and subtract from a partial description? Creating a "full
fidelity infoset" is not so hard, given that XML 1.0 already contains the
required production rules. Who cares if S is present 74 times? That is not
so complicated.

Jonathan Borden
The Open Healthcare Group
http://www.openhealth.org