Having the MARC21 fixed fields, tags and subfields in a database first allowed me to do some quick statistics based on the names of fields and fixed field values. These results are in a page on the Futurelib wiki [4]. I was particularly struck by the large number of different fixed field elements (many of which are probably encoded only rarely in actual instance data), and how many of them have “non-values” like “unknown.” The use of values for “unknown” and “no attempt to code” are directly related to the characteristic of fixed length fields made up of positional data elements, where all positions must be filled in to retain the positioning. It becomes therefore necessary to decide if “unknown” is truly a value or can be treated as the absence of a value. As I once argued at a meeting of the MARC21 standards group, MARBI, about the use of the value “No attempt to code,” it is highly unlikely that someone who is not going to bother to code a value will carefully encode their intention not to bother. Some of these values may exist in instance data only because they have been used as defaults. In the extraction of fixed fields that I have done so far I have ignored the three values “Other,” “No attempt to code,” and “Unknown” although there may be a case to be made for including “Other” in this analysis (and I may decide to add those later). I also did not include elements that have a status of “Obsolete” in the MARC21 documentation. Any of these could be added later if there is a need for them