- From: Hugh Glaser <hugh@glasers.org>
- Date: Sat, 4 Jul 2020 12:35:56 +0100
- To: David Booth <david@dbooth.org>
- Cc: semantic-web@w3.org
Just to add to this. If you embrace the idea that URIs are important (!), and manage their lives in specialist knowledge services (such as we do with sameAs® services), these problems pretty much evaporate. > On 4 Jul 2020, at 02:22, David Booth <david@dbooth.org> wrote: > >> <snip/> > > FWIW I agree with Dan's point, though I don't look at it so much as "using up" the URI space, but rather as a combination of: > > - Creating stable URIs is much more difficult in practice than in theory, because it requires a hosted domain name that will live for a long time. RDF producers should not be forced to bear that up-front cost. This is often said, but not wholly true, I'm afraid. "requires a hosted domain that will live for a long time" - the domain doesn't need to be hosted, can disappear the next day, or even never exist:- the URI will be stable in the sense of not changing "meaning"; it just won't resolve. not resolving is not necessarily a problem - IDs themselves are useful. You, I and others can still make all sorts of statements about the resource - in fact, I suspect that the most common "use" of URIs does not involve resolution. People probably use DBpedia URIs a lot, while only occasionally actually resolving them to get the RDF, perhaps for reconciliation. "The 404 problem is not a problem" and in any case, given a source of sameAs chains with a URI in it where one actually resolves, a "meaning" of that URI, in terms of resolving, can always be established. For example, http://www.uk-postcodes.com/postcode/SO171BJ seems to have died, but using sameAs.org you can get to the Ordnance Survey URI, and thence to RDF. This is true for many, many of the URIs for the early SemWeb, such as data.gov.uk such as school URIs. So all these old URIs are still perfectly serviceable and useful. It's only when someone else (or you subsequently) decide to use that URI for another, and different, resource do we get a problem, especially if a new document gets put at the corresponding URL. And of course, using UUIDs or similar can really help here. URIs in a proper identifier ecosystem have astonishing longevity. The question of security and spoofing etc does come in, of course. But that is actually a wider question of how do you trust *any* URI you resolve - and you better have a policy for that. > > - Developers should be able to use their own local names for things, but map them to well-known URIs later Yes. Absolutely. There are many reasons for this as a good policy. And there should be a good URI management services ecosystem that supports that ;-) Best Hugh > > See also https://github.com/w3c/EasierRDF/issues/17 > > David Booth > -- Hugh 023 8061 5652
Received on Saturday, 4 July 2020 11:36:16 UTC