Rhino, Nashorn and V2, V3 & Java 1.8

[Rod Widdowson]

> >I didn't recall that the data connectors actually exposed IdPAttribute
> >objects, 

They absolutely do:  The IdPAttribute  they go into the common bucket of
resolved attributes which are available throughout the resolution process
BUT (and see below) this is not the bucket which is the output of the
process.

> I was thinking something else was used internally. There's always
> >been this odd disconnect between the data connector outputs and attribute
> >definition outputs in my mind.

It's all rather too complex for my taste, but it has always worked.   I
have, from time to time, understood exactly what it did and convinced myself
that it was OK and for the rest "if it ain't broke don't fix it".  

There is a complex web of attribute definitions, resolved IdPAttributes,
data connectors, resolved attribute definitions and resolved data
connectors, but fundamentally:
 
> is it fair to say that the difference is
> essentially that we don't expose a way to attach encoders to data
> connector output (nor is it obvious how we would, I'm just trying to
> articulate the difference)?

Right.  But also at the end of attribute resolution the result is
constructed by inspecting (only) the resolved attribute definitions which
contain a reference to 'their' IdPAttribute.  So there are many structural
reasons (as well a config ones) why only the output attribute definitions
turns up after resolution.