In addition to focusing on goals, I would like suggest that we study what 40 years of machine-based bibliographic practice tells us. Oh wait, that's already been done: "Implications of MARC Tag Usage on Library Metadata Practices" http://www.oclc.org/research/publications/library/2010/2010-06.pdf One essential fact this report uncovers is that "only 21 to 30 tags occur in 10% or more [WorldCat] records." I would like to suggest that one potential implication of this finding is that we should think about allowing our 40 years of actual practice to identify the core set of elements that clearly are used in bibliographic data, and sequester additional complexity in separate packages that although they can travel with this core they can also be easily ignored by applications that don't choose to address the added complexity. In other words, one of our goals should be "simple requirements should be simple to accomplish". Building mechanisms for various communities to build out richer descriptions for particular kinds of resources is great, but it should not happen at the expense of added complexity to a core description. The monolithic nature of the MARC record has been, I submit, one of its chief problems. When processing it you never know whether you will see a couple dozen data elements or nearly 3,000, so you have to code for every eventuality. Think modular, with a manifest at the top so you know what you have in hand. Roy Tennant These are my own opinions and I do not speak for my employer.