The FRBR Blog

Thursday, the last day of programming at IFLA 2008, was when the Working Group on Aggregates met. Ed O’Neill of OCLC is the chair, and there were six others at the table, including Barbara Tillett and Judy Kuhagen of the Library of Congress. The three of them were the most active in the discussion, though I think Maja Žumer would have spoken up too had she been there. About a dozen people watched, most of whom I recognized from earlier FR* meetings.

There were three handouts for group members, and, generously, we observers all got a copy. The discussion centred around two things on the agenda, so I’ll just summarize what was said. No decision was reached, so there’s no official answer on how FRBR will handle aggregates yet.

Here are the “data model principles” that had been set out in some earlier discussions to help test the three ways of modelling aggregates;

• Inheritance: Properties (attributes and relationships) are inherited by subordinate entities (children) from superior entities. The properties of a work are inherited by its expressions; the properties of an expression are inherited by its manifestations; and the properties of a manifestation are inherited by its items.
• Universality: If an entity is a work in any of its manifestations, it must be a work in all of its manifestations. [O'Neill explained: the criteria for deciding if something is a work shouldn't depend on its manifestation.]
• Distinctness: A non-aggregate work is the smallest distinct and autonomous entity.

After that they got on to the three models they had been considering:

• Work-of-Parts Model: The aggregate as a whole is a work; individual components are simply parts of the work. [Tillett pointed out that "component" already means "component work" in FRBR and they didn't want to confuse the terminology.]
• Manifestation-of-Works Model: The manifestation is an aggregate of works and may include an “aggregating” work.
• Work-of-Works Model: Aggregates are works that are comprised of other works.

No-one had any other models to consider, and nobody there thought the first one was valid, so it came down to Manifestation-of-Works (with O’Neill as main advocate) vs. Work-of-Works (with Tillett as main advocate).

A few points from the discussion:

• Tillett said inheritance holds down through Work, Expression, Manifestation, Item, but there’s no inheritance between Works in the Work-of-Works model. There is a Work-to-Work relation (whole-part, sequential, etc.) between them, with no inheritance happening.
• There’s a difference between an aggregate that’s an augmentation, for example adding an introduction to a novel, and one that’s a collection, for example putting the three novels of a trilogy together into one book. Kuhagen made the distinction between pre-formed and post-formed collections, where the difference is when the decision to make the collection was made.
• But none of that matters for this discussion, Tillett said, and she said something she repeated a few times through the meeting: FRBR is meant to work at a high conceptual level, and not get into applications. (Applications as in “the model applied to this situation,” not application as in “computer software.”) FRBR should only think about aggregates in general, any kind of aggregate, and not worry what type an aggregate is. That level of detail belongs in applications of the model, where people can make their own local rules and interpretations.
• O’Neill said there would probably eventually be a subclassing of aggregates, because they are not all the same. This could be done in a FRBRoo way.
• What if in different applications of FRBR there are different rules about what a work is? O’Neill asked. How can we share information if we don’t agree on what a work is? “We can link it,” Tillett said. She agreed that something could be a work in one application but not in another, though theoretically it really is a work.
• O’Neill said that many things are works don’t need to be recognized as such, for example a very short introduction to a book. But that’s a local implementation decision, and if the introduction were to be recognized, it would have to be recognized as a work of its own. Everyone agreed on that. The universality rule was reworded to use “can be” instead of must: “If an entity is a work in any of its manifestations, it can be a work in all of its manifestations.”
• There was some discussion about the distinctness principle and whether it begged the question of an aggregate being a work of works. They all went to a draft report and looked at something and clarified that a list was ORed, not ANDed.
• In the end only the distinctness criteria was seen as useful: inheritance doesn’t apply to Work-of-Works and universality as modified is a given so it’s not useful as a test.
• They got into a metaphor next to think about an “aggregating work.” Take Patrick Le Boeuf’s special FRBR issue of Cataloging & Classification Quarterly, which contained articles by many people. It was also released as a book. O’Neill said each chapter is an independent work, but Le Boeuf’s role as an editor is that of an aggregator. Each article is like a brick, and Le Boeuf’s intellectual contribution was to make the mortar that holds all of the bricks together into a new work. That mortar is an “aggregating work.” The aggregate work is the whole wall of bricks and mortar.
• Tillett said there was no need for this. The articles and the books are all just works, and they are related to each other. That’s all.
• Gordon Dunsire spoke up from the audience with an objection to this “aggregating work” idea, I think saying that if you take three bricks and put them together with mortar and call that new thing a brick, then you have bricks containing bricks, and it gets recursive and you get into problems. You can’t have a brick containing other bricks. (This reminded me of Russell’s paradox.)

Eventually they decided on a bake-off. They would take some examples and represent them using both the Manifestation-of-Works model (O’Neill and Žumer) and the Work-of-Works model (Tillett and Kuhagen). They would look at those and decide what to recommend as the amendment to FRBR. The examples will be: a collection of music by Sibelius (based on an example Eeva Murtomaa brought), an augmented version of Humphry Clinker (with an introduction, illustrations, that kind of thing), Le Boeuf’s FBBR journal issue/book, and a moving image to be decided (I’d guess a DVD with extras). In order to have this all done by next year’s conference in Milan they need to have the recommendation done for February, so they’re going to get the modelling done before October.

This working group meeting was really lively and fun to watch. I’m glad I went.