support citations for datasets

[twhiteaker]

In EML 2.2, datasets can have these child elements:

• referencePublication - (0, 1) THE publication that describes the dataset, or when a paper is intended to be the canonical or examplar reference to the dataset
• usageCitation - (0, many) a publication in which the dataset is used
• literatureCited - (0, many) a publication referenced in the dataset or its metadata, e.g., a paper on a particular method used to collect the data

To support this, the dataset table could be related to publications. Here are a couple of ideas:

(a). Three new tables, one for each type of publication listed above.
(b). One table, with a column, RelationToDataset with possible values referencePublication, usageCitation, literatureCited.

(b) seems more efficient, but I don't know how you'd support (0,1) for referencePublication and (0, many) for the others.

The citation itself could be a varchar field storing XML or BibTeX. A separate column indicates type of content.

[gastil]

(c) A cross-reference table

DataSetID integer NOT NULL,
pub_id varchar NOT NULL,
relationship varchar NOT NULL,
notes text,
CONSTRAINT data_pub_xref_pk PRIMARY KEY ("DataSetID",pub_id,relationship)

• I include relationship in the pk because, theoretically, a paper could appear in more than one of those elements. Not likely, and I will entertain argument to remove that from pk.
• All three of those pk columns must be a controlled vocab. The relationships can be in a CHECK constraint for the three elements you mentioned, plus 'undefined' for wiggle room. Obviously the parent for DataSetID table DataSet. The parent for pub_id will be whatever that site uses locally. That does present an install challenge.
• Because we want to keep our core-mb at 4th normal form.
• Because the biblio schema already supports pub_id (although I forget the name of that column, but you get the idea.) A site may have another schema for pubs, as MCR does, but still want to use that schema together with core-mb.
• How the citation text string or pieces are stored is outside the scope of core-mb and within scope of a biblio schema. Ideally that biblio schema is within the same database as core-mb so it is accessible to VIEWs and and FKs.

[gastil]

Side comment: usageCitation within the EML is not how I want to manage that information. External tools allow us to add such relationships after a revision of a dataset gets its doi minted. You can see how, otherwise, one has a chicken-and-egg conundrum.

[twhiteaker]

I see your point about usageCitation.

How would you enforce (0, 1) between a dataset and a publication for referencePublication?