XmlSchema Object Model, Post-compilation InfoSet (PCI) and API consistency

Note: this entry has moved.

The Schema Object Model (SOM), much like the DOM, holds pre- and post-compilation (validation for DOM) information items. However, the model is not consistent for both, in the case of the SOM. For example, for the pre-compilation phase, you iterate the top-level items through the XmlSchema.Items collection:

foreach (XmlSchemaObject item in xsd.Items)

On the other hand, after schema compilation, when all the items in that "cat bag" go to the corresponding XmlSchema.Attributes, XmlSchema.Elements and XmlSchema.SchemaTypes properties, you can't keep iterating like that because now the iterator returned by the XmlSchemaObjectTable.GetEnumerator (they are all of that type), returns an IDictionaryEnumerator, meaning you now have to use either:

foreach (DictionaryEntry entry in xsd.Elements)
{
    XmlSchemaElement el = (XmlSchemaElement) entry.Value;
    //...
}

foreach (XmlSchemaElement el in xsd.Elements.Values)

Which isn't very consistent with the approach for pre-compilation equivalent property. Besides, it's always a bit cumbersome having to deal with two sets of available properties for the pre/post state.
I'm wondering whether it would be better to have a separate object model extending the pre- one and adding the new post- ones. I think a cool API design would be to turn all of the pre- SOM to interfaces, and make the concrete pre/post implement it. The later could implement them explicitly, so that by default you don't have to "see" the pre- when you're dealing with the post- SOM.
Any thoughts?