This is needsAgreement because adding that statement would conflict with existing text in 3.4.3 SML reference schemes that are not SML-IF inter-document references saying "Third, when creating a new sml:ref scheme, authors MUST be explicit about whether the scheme is an SML-IF inter-document reference."

Less obviously, the arguments around saying EPRs are not SMLIF idrs are valid for the SML-defined EPR scheme, and for "random" EPR content, but would not necessarily be valid for every reference scheme.  If the reference scheme constrains things sufficiently to define exactly how its references are resolved, there would be no problem doing so in an SML-IF consumer.  Just because we have not defined such a scheme yet is no cause to preclude its future definition.

It would also exclude the case where the consumer has out of band knowledge about how to resolve EPRs (say a run-time option).

If we feel the need to say something about "naked" EPRs, i.e. those not recognized as part of a reference scheme instance, then "MAY interpret as idrs" would be the strongest we could say.  I have no problem doing that, but that is a very different semantic than what the summary proposed.  Whether or not we decide to make that stmt however, it would be inappropriate to me to have SML-IF constrain what reference schemes (spec'd in SML) are allowed to define as their syntax and semantics.  SML-IF could validly layer on addl requirements wrt SML, but I don't think it should be removing freedom granted in SML who "owns" the def of and reqts on reference scheme definitions.

Resolution
1. Remove last sentence in 3.4.2
2. reword 2nd to last paragraph in 3.4.2 1st sentence to say that EPR scheme is not an inter-document reference because the scheme is defined that way.
3. Add statement in EPR scheme definition in SML spec to say that the EPR scheme is not an inter-document reference in SML-IF.
4. Check rest of section to tighten up wording based on new requirement that scheme definition must state whether or not it is an SML-IF inter-document reference.
5. Shorten 1st sentence of last para reads "..hints as inter-document references."

EPR Scheme definition changed to:
========
The EPR reference scheme MUST be implemented by using an instance of the wsa:EndpointReference global element declaration [WS-Addressing Core] as a  child element of the SML reference element. Instances of the EPR reference scheme MUST NOT be interpreted as inter-document references in the context of an SML-IF document.
=========


Section 3.4.2 has been changed to:
============
3.4.2 Definition

In the context of SML-IF, an inter-document reference is defined as:

   1. Any content in a document in the interchange set whose type is xs:anyURI or a type derived from xs:anyURI.
   2. Any SML reference represented using a reference scheme whose definition specifies that it is to be treated as an inter-document reference in the context of SML-IF.

For example, an xsi:schemaLocation attribute is defined to be of type list of xs:anyURI. These come in pairs, one for the namespace name, and one for a hint as to the location of a schema document defining names for that namespace name. This makes the "hint" URIs in xsi:schemaLocation attributes SML-IF inter-document references.

xsi:schemaLocation hints MUST interpreted as inter-document references.

Similarly, an sml:uri element that is contained in an SML reference is an inter-document reference because the definition of the URI reference scheme states that it must be interpreted as an inter-document reference in an SML-IF document. In contrast, the wsa:address element in the the Endpoint Reference (EPR) reference scheme is not an SML-IF inter-document reference because it is not specified as such in the definition of the EPR reference scheme. 
===========

I agree with the change described in comment# 3.