- From: IS4 <is4.site@gmail.com>
- Date: Sun, 31 Dec 2023 15:25:54 +0100
- To: semantic-web@w3.org
- Cc: gk@ninebynine.org
Hello! I am sorry that I missed this message and could not answer earlier. I am happy that my scheme was noticed and I will gladly answer any further questions. > It looks to me like yet another attempt at a persistent identifier scheme. The original intent was to use lid: as a sort of SPARQL compression, removing the need to write a query each time when you look for a resource with a particular (non-URI) identifier. With a particular resolver (such as the one of mine at https://data.is4.site/lid/ ), a very short URI can be written "by hand" and expanded to a link to a SPARQL endpoint (see https://data.is4.site/lid://dbpedia.org/rdfs:label/Earth@en for an example, or https://data.is4.site/lid://dbpedia.org/rdfs:label/Earth@en?_action=debug for the generated query). Whether that constitutes a persistent identifier depends of many aspects ‒ whether the endpoint actually stores persistent data, whether the property chain consists only of owl:InverseFunctionalProperty (implied at least), and whether a concrete literal (plain or with a datatype/language) or a range of literals (untyped or a language range) is used. Even the prefixes are somewhat volatile (the target's knowledge of a prefix can be used). Therefore, a URI in the form <lid://server/a:b/c> is as persistent as the server's knowledge of a:, the property a:b, and the particular entity identified by that property as "c", and, from a semantical point of view, it identifies anything in general that satisfies these conditions, broadly. In that sense, it serves only as a shortcut to the real entity ‒ when you are describing an external dataset in RDF, it is better to stick to its particular identifier scheme when possible, and in your own dataset, it would be better to use http: since lid: URIs are somewhat self-referential in that case. That being said, throughout the process some (in my opinion) cool possibilities arose naturally from the syntax, when you omit the host. This allows you to universally identify literals, for example <lid:15@xsd:decimal>; you can "shorten" vocabularies, e.g. <lid:uri/$foaf:> (identifying any vocabulary that is associated with this prefix), and you can definitely use existing properties to form persistent identifiers (but not locators), such as <lid:schema:isbn/9780345339706> (or anything else that is not a URN namespace yet). Feel free to use these for RDF purposes. ;-) Sincerely, IS4.
Received on Thursday, 4 January 2024 03:37:39 UTC