Automatic recovery from failures of messages within a data interchange

Data interchanges are processed so that problematic individual elements within the data interchanges do not cause the processing of the data interchanges to be suspended unless there is a catastrophic error. The data interchanges are split up by flat file and/or XML disassemblers of a message engine into independent messages even if some elements are incorrect or cause processing failures. Those messages in the interchange that may be processed are processed, while those that would cause processing errors are suspended in a suspend queue for recovery processing. Even when a message is placed in the suspend queue, processing continues with the next independent message in the data interchange. Messages that are placed in the suspend queue for lack of routing information may be resumed by providing the routing information and again attempting to route the previously suspended message. The method of the invention is scalable since it works regardless of the size of the interchange or its individual message elements.