The University of Washington conducted an experiment by creating an RDA input form with output in multiple RDF schemas. One goal, among others, was to demonstrate that RDA cataloging (input) can be easily output in multiple schemas using a processing pipeline and mappings.
This experiment was successful and output was achieved in RDA/RDF and BIBFRAME. The key is to use an intermediate format that is as fine-grained as the most detailed output, in this case RDA. We also showed that output in these schemas can be generated in an automated fashion using a pipeline.
The implication for future production cataloging systems is that input and output should not be directly tied to each other. Rather, a cataloging system should have sufficient flexibility to output in multiple schemas, which can be achieved in an automated way. This will facilitate the exchange of data at the level of granularity needed by different audiences.
In a linked data world, the technical environment is fundamentally different. In MARC, cataloging data are directly encoded in the carrier; in fact, part of a cataloger’s work is to encode the cataloging. Also, there is only one carrier. With linked data, there are more options: cataloging and carrier can and should be separate. Catalogers should focus on the creation of cataloging data and do not need to encode them in a schema. This can be done automatically in the background. Also, multiple schemas are easily used, and should be.
For more details on the experiment, see http://www.lib.washington.edu/msd/pubcat/ld/input-form