[TSC-public] Filing Sources

Thomas Wetmore ttw4 at verizon.net
Mon Nov 3 12:21:48 CST 2014


At the risk of sounding like a broken record once again, the whole source thing can be modeled very easily. It takes care of all war stories and examples and approaches that are being bandied about. I have been explaining this approach for more than twenty years. It is so easy, basic and obvious, it should make you fall off your couch.

1. A source record can represent a whole Source, or a part of a Source, or a part of a part of a Source, and so on and so on, depending upon the predilections of either software or website or user or data provider or published convention or archive policy and so on and so on.

["Source" in upper-case means a real source in the real world, whereas "source record" means a representation of that Source or a part of that Source in computerese.]

2. When a source record represents a part of a Source, it should have a source reference to a containing source record that represents the full (or next higher level part of the) Source. When a source record represents a full Source is does not have a source reference to anything higher. So source records can be arranged in hierarchical trees, but they don't have to be. Totally up to you and up to data providers. Do it your way, and be comfortable in the rightness of your being. If you can't stand the idea of hierarchical sources, don't use them.

3. Any other record in a database that holds or refers to information extracted from a Source (my favorite example is the persona record, but it doesn't have to be), uses a source reference to the source record that best represents its Source in the real world.

4. A source record has a type (e.g., book, journal, journal-number, journal-article, website, webpage, census, census-page, and up the kazoo).

5. Each source record type has lots of possible "metadata" (which are what are called properties or assertions in other record types), and our model must handle all the possibilities, probably with some extension capabilities.

6. This can be handled with TRIVIAL extensions to GEDCOM. a) Add more properties (level 1 lines) to the "0 @X@ SOUR" record, including a "1 TYPE" line; and b) add a "1 SOUR @X@" line to the "0 @X@ SOUR" record to allow the hierarchy.

7. Citations can be generated by following up the source chain (which only has one link when the source record represents the full Source), extracting properties ("metadata") along the way, and using that metadata to construct the citation. One pattern can be followed for footnotes, another for bibliographic entries and another for end notes. Use another set of templates if you want another style.

This would take care of every example discussed lately. But not only every example, but every way anyone would want to handle every example. Do it your way. If you want one source per census, you got it. If you want one source per page per census, you got it. If you want to treat a page as a source within a census as a higher level source, you got it.

Tom Wetmore


More information about the TSC-public mailing list