role, arcrole, purpose, nature

[Tim Bray <tbray@textuality.com>]

I just now got around to reading the rather good discussion of this
stuff as related to RDDL.  In particular the discussion about "purpose" 
and "nature".  The fact that XLink happens to provide two different
role attributes seems to map nicely onto this, and having closely
reread what the XLink draft says, the language is fortunately loose
enough that we could assign these things either way I think without
doing violence to the spirit of XLink.

I could live with either way of doing it.  I'd argue for using role=
for "purpose" and arcrole="nature", just because I think that in this
type of application, purpose is more important than nature.

For example, I'm trying to think of application scenarios where I'd
charge off and run a schema over some instance when the only knowledge
that I have is that it's a schema.  I just can't see it; for all 
I know this is a schema that provides heavy type-checking of <xyz> 
element content and skips lightly over everything else.  Same comment
for CSS stylesheets, DTDs, java classes, you name it.  At the end of
the day I have never bought into the idea that a vocabulary must
have a special "definitive" or "canonical" schema; maybe I'm in a 
minority here.

I still think the idea of having a rddl.org/arcroles.html resource is
a really good one, except we might want to call it natures.html.  
Even if any sensible person can see what the identifier for XSD files 
has to be, it's a good idea to have that written down somewhere.  And 
there are lots of others where the identifer is just going to be 
somewhat arbitrary, or even where it's not arbitrary but also not 
obvious.

Once we define a minimal RDDL on this basis, the next interesting
question is: can we say anything definitive about a canonical list
of "purpose" values?  While this sounds like something that could
escalate into a huge task, I can think of lots of things that would
be nice to be able to specify:

- schema designed to support authoring applications
- stylesheet designed for browser X version Y and browser Z version W
- processor that makes RTF
- executable code that performs spell-checking

Note that all the above "purpose" could orthogonally be combined with the
"nature" values.

Hey Jonathan, waiting for the next RDDL draft reflecting all this
useful discussion.   Or is it there now? -Tim