Re: error handling of dangling references in encoding

+1 for SHOULD

Christopher Ferris wrote:

> On the call, I had suggested that use of the RFC2119
> SHOULD qualifier might be used in our spec. This would
> provide guidance to implementers that leaned towards
> MUST, but at the same time did not impose a hard requirement
> on the receiving node.
>
> As I understood from the discussion on the call, imposing
> a requirement to return a SOAP Fault might be problematic
> in cases where there might be good reason not to have to
> parse the entire message to determine if there were dangling
> references.