DOCTYPE, Namespaces, Schemas

Building chains and gateways requires identification You can throw documents at a transformation and let the successes appear while the failures disappear, but sometimes its useful to figure out which document goes to which transformation.
Multiple identifiers in documents Documents may have DOCTYPE declarations (including identifying root element names and system and public identifiers), plus schema location identifiers, namespace identifiers, and maybe even MIME types.
Clouds In some ways there's just too much information. Transformation architectures may need to build identification architectures as a front-end.