Danger, Will Robinson!
Author Identifiers OL\dA are often non-unique. Several such OL records may exist for a single author, even with the exact same spelling of the name. In other cases, one OL record may reflect many authors with similar names. OL has been dragging its feet for years on implementing consistent authority controls. I've tried to get their attention on this problem to no avail. We'll need to be able to deal with this. Work Identifiers OL\dW are often non-unique, with the same or varying title, linked to various author records. Cleaning up the mess at https://openlibrary.org/search?q=Odyssey&author_key=OL6848355A could take weeks. Edition Identifiers OL\dM are even more of a mess. They conflate different editions under one record, and have multiple records for one edition. All this said, however, OL remains one of the most useful ways to locate a freely-readable copy of many sources. As such, OL identifiers are given for editions or works cited, per w:en:SAYWHEREYOUGOTIT. We'll need to accommodate the multiple OL identifiers attributed to one work, edition, or author. We'll also need to distinguish the multiple works, editions, or authors conflated by one OL identifier. It's nontrivial. It may be impossible to code. Ideas? LeadSongDog (talk) 17:30, 16 June 2016 (UTC)
- @LeadSongDog: Thanks for the info! For specificity, can you give examples of such duplications for A, W, M? (I haven't yet seen a M ID, I'd like to see one).
- Authors with many OL ID, OL ID mapping to several authors (first one is a mistake on WD)
- IMHO WD only needs to worry about the multiplicity if OL goes ahead and removes duplicate IDs without keeping a redirect to the surviving (merged-to) ID: do you have evidence of this happening? Otherwise WD could help OL by finding such duplicates (as constraint violations) but WD can't clean up OL's database. If you yourself say OL IDs are useful, we should keep them. --Vladimir Alexiev (talk) 10:07, 7 June 2017 (UTC)
- @Vladimir Alexiev: Well, an obvious case is OL2624944A. The appropriate quick intervention (which I did in this case) is to suffix " (undifferentiated)" to the author name. The harder job is to mine each of the linked works and its editions in order to find sourcing for more completely spelled names (often via an edition's ISBN or OCLCno), then edit the OL\d*W record to reflect that finding. In most cases there will already be an OL\d*A record to link in. Sometimes there are duplicates even at the better (more specific) spelling.  and  are the same person, while  is quite another. VIAF and ISNI can help to clarify these cases. The latter died before the former was born. LeadSongDog (talk) 17:06, 7 June 2017 (UTC)
Split this property
This property is currently used for the OpenLibrary identifiers for works, editions and people. But, sadly Open Library is using different URI patterns for each of these type (
https://openlibrary.org/authors/$1...). It would be nice to split this property in 3 to avoid this problem and to be able to have better constraints. I would suggest to keep this property for the author identifiers (around 119117 usages) and create two new properties, one for works (around 956 usages) and one for editions (around 34324 usages). What do you think about it? Tpt (talk) 14:15, 13 June 2018 (UTC)
- I think a feature was requested to address the uri question. It seems to be in the works somewhere. If the existing property is re-purposed as suggested, users might get incoherent results.
--- Jura 14:19, 13 June 2018 (UTC)