OpenSAML user discussion

["Scott Cantor" <>]

Brent Putman wrote on 2009-02-09:
> Right, since AttributeValue is defined as an xs:anyType, we don't supply a
> concrete implementation of AttributeValue.

I do, but its implementation is nothing more than an ElementProxy (my XSAny) 
that defaults the element name.
 
> Either you
> write one, or you could use an XSAny that supports an open content model of
> any attribute and any child element.

Which is where we sort of diverge, I think, because I consider that a 
natural/logical thing to do and would actually document that as the solution 
for most people to use for simple cases. Type safety has value, but so does 
expedience. What matters is that the XML is handled safely and that 
signatures hold. If it weren't for signatures and encryption, none of this 
code would exist, in fact, because it's too thin a wrapper around the DOM to 
add value otherwise.

> I would say, though, that if you do have a formal schema and type model,
> etc, that you probably want to  go ahead and implement providers for them.
> At least in my head, that lends itself to more conceptual correctness.  I
> mean, we could use XSAny for just about everything at a functional level,
> but we don't.  It's mostly intended to support unmarshalling unknown types
> that you receive, and provide access to the content model in an open
> fashion.  It's not really for building and marshalling things in your own
> code (although it can function that way).

The problem is that with things like AttributeValue, where use of xsi:type is 
discouraged, you *will* (I think) get an XSAny on the way back in. So the 
cost of creating the code has to be balanced by that.

-- Scott