Ariba OnRamp Large Document Handling

We’re going to be using Ariba OnRamp to sit in front of Trading Networks and recieve incoming requests containing file attachments. I’ve reviewed the large document handling documentation that came with Trading Networks, but what about Ariba? The OnRamp wm.b2b.cxml:receiveCXML service will parse the MIME/multipart message into a cXML node object and attachmentWrapperCollection, which are then passed to Trading Networks via the pipeline. What happens if the attachments are very large? How does Ariba handle this?