metacpan
Make metadata available as Linked Data
I mentioned this briefly on IRC, and @oalders encouraged me to submit it here.
Some years ago, @tobyink made all of CPAN available as Linked Data with RDF using an old version of the MetaCPAN API. Since then, it has stopped operating, and while @tobyink may redo it, I'm thinking it might need to be done slightly differently, and based on a new framework.
Some context: I'm creating a framework where tests can be formulated in RDF for the Solid Project, where test reports can also be returned as RDF. Linked Data has a basic principle that you give everything a URL to identify it, and then you return some RDF if someone GETs it. For this, I need an RDF URL for everything that goes into those tests, and that's a bunch of CPAN stuff. We are also formulating module metadata as RDF using the DOAP vocabulary, which means that not only can we have a consistent presentation across projects, we can also store all this and query and analyze it. We can also link to Debian and Git, which also exposes RDF.
Once done, I figured it makes sense to have this as part of MetaCPAN rather than a standalone project like @tobyink 's were.
For this, I need URLs for
- authors
- distributions
- individual releases
- individual modules
So, the question is then how the URLs should look. Perhaps it would make sense to write them with ld in the URL, for example https://metacpan.org/ld/?
The author URL would straigthforwardly be
https://metacpan.org/ld/author/$pause_id
I think that
https://metacpan.org/ld/distribution/$dist
could reasonably always refer to metadata for the distribution as of the latest version, where it has links to versions with their changelogs at
https://metacpan.org/ld/distribution/$dist/$version
I'm a bit conflicted on whether the list of modules should have versions as well, and if they should perhaps live under their distribution. I don't have a concrete use case for that right now, so I'm think that perhaps only the latest release needs to have RDF metadata. So, perhaps
https://metacpan.org/ld/module/$module
would suffice.
Do you think this is something MetaCPAN could have? If so, I might get around to contribute the code.
My version is working again. I think what's wrong is it's often running out of memory when it rebuilds its database. The VPS it's hosted on is really old. I should move hosting some time; there are much cheaper and better hosting packages available these days.
Great stuff, @tobyink !
Perhaps we could host it on my box, and perhaps collaborate on it?
