James - I support this 1000%. An important idea, and namespaces for queries are very valuable and currently underdeveloped + underappreciated in many environments.
I would want the query-space to be editable, with a version history for a given query ID: the idea being to make it easy to create and refine a sharable view of knowledge. Which is what we do now when creating a new page, portal, entity, category... To your last concern, which I share: -- I would not auto-assign these IDs, but have them easy to request; in the fashion of bit.ly -- on request you are offered a (not human-readable) ID for that query, and can ask for a specific short string instead. >From an API perspective, -- I like the idea of a URL that refers to a query and allows one to access its query string / history / last update time / results -- Gerard: yes, as of the [current time] when it is viewed. That's often what I want. It would be somewhat useful to have a way to indicate "as of <timestamp>", and Memento offers a mechanism for this. But we don't use Memento even for web-citations; we just include a "retrieved on" date in the citation data. So: not a priority. SJ On Thu, Feb 18, 2021 at 8:34 AM James Heald <[email protected]> wrote: > Dear wikidata list, > > One of the key things we do as Wikidata people is go round the internet, > hassling people to create nice identifiers for their things, with URIs > and landing-pages that we can link to. > > It brought me up quite short to realise that actually applies to *us* > too -- there is an important thing of ours that haven't got a linkable > identifier for, that it would be useful if there was a short linkable > url for, and that is WDQS queries. > > > So here's the use-case: > > [[User:PKM]] and I have been working with a new external project who > want to build a "Gazetteer of Early Modern England and Wales" (EMEW) -- > basically a historical GIS for 15th & 16th century England and Wales, > able to plot things on this map: > https://viaeregiae.org/index.php/map/?layers=l9001l0007 > > There's huge scope for collaboration with Wikidata, with deep linking > both ways, as we both try to improve our coverage of C16-C17 England and > Wales (expect WikiProject [[:d:WD:EMEW]] just as soon as we can get the > pages made) > > Something that we realise we want to be able to do, from the EMEW site, > is for a user to be able to give it the URL for a WDQS sparql query, for > the site to send that to WDQS, get back a file of EMEW ids, and show the > results on the EMEW map. > > > Now of course that *can* be done with full WDQS query URLs. But they > are horribly long and awkward. > > What would be much nicer would be if each time WDQS ran a query it > hadn't seen before, it generated a new short identifier, that the UI > would display, and which could then be used to refer to the query. > > So on the EMEW site, one would just put in the short identifier, that > would send it to WDQS with some appropriate URL-start to show it was a > request for data rather than for a link, back would come the results -- > and all the user would have had to copy in was a short identifier. > > > Of course to some extent the URL-shortener does this, but there are some > issues: > > 1) The maintainers point-blank refuse to let it allow URLs of more than > 2000 characters. ( https://phabricator.wikimedia.org/T220703 ) Gnarly > WDQS queries can often be longer than this, sometimes a lot longer. > > 2) The short URL could be for anything on any wiki site -- the EMEW site > can't be sure that it corresponds to a SPARQL query > > 3) The short URL needs to be adjusted, to turn it from a WDQS url that's > a link to the query in the GUI into a WDQS url that's an external > request for query results. This is not straightforward. > > > A short identifier for a WDQS query would get round all these things. > > It also might be one step forwards towards creating a place like Quarry > (https://quarry.wmflabs.org/) where users could save their queries, > share them, document them, see other people's shared queries, and come > back to them later. But that's another ticket > (https://phabricator.wikimedia.org/T104762 open since July 2015). > > All I am suggesting, first, is an identifier. > > > One objection that I thought of might be that if identifiers were > automatically assigned, without having to actually request them, then > people might be able to "spy" on what other queries people happened to > be writing at any one time. I don't know how serious an objection this > is - it doesn't seem to be a problem for Quarry - but could largely be > avoided if the query-number was hashed to make the sequence less > predictable. > > (Or alternatively, query-numbers could just be issued on request). > > > Anyway, just putting this out here, for thoughts. > > Best wishes to everybody, > > James. > > > > > > > _______________________________________________ > Wikidata mailing list > [email protected] > https://lists.wikimedia.org/mailman/listinfo/wikidata > -- Samuel Klein @metasj w:user:sj +1 617 529 4266
_______________________________________________ Wikidata mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wikidata
