What’s a CURIE, and Why You Should be Using Them | Biopragmatics

[utterances-bot]

What’s a CURIE, and Why You Should be Using Them | Biopragmatics

Compact uniform resource identifiers, or CURIEs, are an important formalism for referencing biomedical entities. This post explains what they are, how to write them yourself, and a brief outline of how they fit in to the semantic web, linked open data, and open biomedical ontology worlds.

https://cthoyt.com/2021/09/14/curies.html

[mpo-vliz]

This feels unnatural to somebody with a semantic web view -- the URI is the semantic sound and verifiable thing, the prefix, in contrast, is only of temporary and local value.

That view prevails when we mark this section in the text

Since everyone agrees on what ChEBI is, it makes sense to ...

as the most dangerous (unverified claim) part of it.

I get it, this wants to be 'bio-pragmatic' - but this has the red-flag-implicit-assumption written all over it. And you know the saying: to assume is to ass-u-me: it makes an ass out of you and me?

Point is: that statement is very likely only true at this day and age, and only holds for some known inner circle of people...

Getting beyond that valid-in-local-use-only would require a central registry for all prefixes to be verifyable in what they actually stand for. Something like prefix.cc or the iana-urn-namespaces registers. All of which are probably equally unexpcected and obscure to everybody else then the bioregistry ?

Hey, this reminds me. (irony ahead) That registry actually even looks like what domain-registrations are doing - a piece of work that ends up being present and leading inside any URI. Well that should ensure semantic clarity for any of them, no? (end of irony)

But as you point out: it did not avoid the mess we all see and have to endure.
But then again, some "pragmatics" tells me that if human-kind can make a mess out of its uri-management on top of a nice domain-registration system, they will surely be able to do it based on these-prefix-is-mine-assumptions.

As Douglas Adams concluded a long time ago #peopleareaproblem

Joking aside though. The root cause of the matter is that of independent invention. Not a bad thing (well not as bad as sloppy coordination which is hard to distinguish). Anyway, it forces to accept the fact that quite simply "aliases will exist" and will further lead to the dreadful "mappings will be needed", further resulting into "wasted time in needless confusion" and "technical scaling and performance limits" will be experienced.

The thing I would hope to eventually resolve those is to introduce a post-factum re-alignment of "canonicalisation" -- where each of the aliasses can be probed to describe their accepted "canonic replacement". If we would just stick to actual URI that can be resolved, than this canonicalisation could even happen online. Visiting the resource can yield a Link header with one of the known link relations to either directly rel=canonical resolve the issue or do it via a mapping resource it shares through RDF at rel=describedby.

Yeah I know, that floats only on the dream that we can actually make all out there to actually agree on some ONE thing at least: a common way to express how we acknowledge we discovered earlier disagreement. Allow me that dream, we all need it, cause it also justifies yours: The really good news is that this possible future actually gives the here-and-now blessing to all more independent local, temporal and pragmatic solutions like the bioregistry. Some day!