[Frederik Baumgardt]

Gary

1. Reagan mentioning types of collections: Digital Collection, Data series, Sensor-based data series, Data set: discrete set, Data stream
2. —
3. PIDs are not substrate
4. need ontology/concepts/mereology
5. —
6. PID != collection/resource
7. PIDs important for RDA
8. identifiers are usually digital identifiers in RDA context
9. identifiers are bitstrings that can uniquely identify a resource
10. identifiers are a property of the resource
11. in RDA identifiers might be related to registries and repositories

Ulrich

1. Collection: PID referencing a digital object containing a list of PIDs referencing objects and pointers to metadata
2. Collection defined by explicit enumeration or by implicit rules
3. Collections can be recursive
4. Collection is finite if the recursively contained PIDs/IDs are finite
5. Finite Collections solve issues in generative sets
6. —
7. Rules:
8. sub-collections (filtered)
9. power collections
10. flattened collections
11. sub-trees in recursive collections [my addition: paths in RC]
12. power collection of all previous versions
13. —
14. Need solid formal foundation for provability
15. alternatively no proof and use other mechanisms to guarantee consistency
16. No decision on whether to use PID, ID or both
17. —
18. currently no predicate other than partOf
19. suggestion to type the partOf relation in collection metadata
20. 2 collections are not the same if elements same, but relation-type different

Tobias

1. Unordered sets can be a benefit over lists computationally (Distributed HashMaps, NoSQL)

Jeremy

1. Link to HTRC collection framework

Keith

1. Collection != PID: PID points to an object that is the collection
2. Recursive collections are limited by their hierarchical (DAG) topology:
3. Let’s not use abstractions that lose semantics
4. —
5. Identifier != Locator

Juha

1. DC Collections definition: just aggregation of items, richer def. would limit applications
2. Collection!=PID
3. Choice of identifier type not trivial

Jacob

1. PID != Collection => referer != referent
2. are collections blank nodes? precision imprecise!
3. set::scalar collection::vector
4. gatheredInto describes process, ignoring possible actors (river/beach examples)
5. role of a collection affects modelling needs
6. use of metadata vs inspection of content – to identify a collection
7. —
8. Identifiers are like labels, names except they’re digital
9. “Identifier != Locator” is not as clear, identifiers are used to locate
10. uniqueness and persistence of identifiers are contigent properties, contingent metaproperties of resources

Harrison

1. need licensing, policy metadata
2. and access control

Discussion