Interesting. Why would you have the prefix "https://machine.wikipedia.org/". This could be useful for many projects particular Wikidata, - not just wikipedia. Would "https://wikimachine.org/" or something be better? "mw" is used as prefix for https://www.mediawiki.org and mw2sparql https://www.mediawiki.org/wiki/MW2SPARQL "mw:P2<t0, t1, t2>" does not seem to be supported by SPARQL and < and > are not allowed in IRI_REF. Ordinary parentheses seems to be allowed. The rule is '<' ([^<>"{}|^`\]-[#x00-#x20])* '>' "mw:P1(arg0, arg1, arg2)" and "mw:P2<t0, t1, t2>" Why are there two kinds of input? Wouldn't one suffice. "mw:P1" Why are the identifiers prefixed with P? That collides with the Wikidata properties. "mw:P1(arg0, arg1, arg2)" This form does not allow for optional arguments. There is also the question of what should be returned. There should be room for error messaging. One idea I have not followed through is to make an empty SPARQL endpoint that would just provide SPARQL functions (extension functions). This way the function could possibly be used in a federated query. best regards Finn Årup Nielsen On 10/17/18 11:32 AM, Adam Sobieski wrote:

Finn Årup Nielsen, Thank you. The techniques could be of use for a number of projects; I chose http://machine.wikipedia.org for brainstorming purposes. I chose the example prefix “mw” for “machine Wikipedia” but we could use another prefix. Similarly with the placeholders for P1, P2. My initial thoughts with respect to optional arguments and variadic predicates are that URL-addressable syntaxes support variadic predicates if a system is such that predicates can be defined with indefinite arity with respect to their T0, T1, T2 and/or A0, A1, A2 parameters. With respect to what is returned, my initial thinking is that the output varies based on HTTP request headers, in particular Accept and Accept-Language . For the Accept values of “text/html” or “application/xhtml+xml”, the content returned could be human-readable wiki content. Other interesting formats to consider include “application/rdf+xml”. For predicates without arguments, e.g. for URLs resembling http://machine.wikipedia.org/P1 and http://machine.wikipedia.org/P2?T0=t0&T1=t1&T2=t2 , the definitions of predicates could be accessed or outputted. With respect to error messages, in addition to the HTTP return codes of 4xx and 5xx, some requested formats may be able to encode error-related data. Best regards, Adam Sobieski ________________________________ From: Wiki-research-l &lt;wiki-research-l-bounces(a)lists.wikimedia.org&gt; on behalf of fn(a)imm.dtu.dk &lt;fn(a)imm.dtu.dk&gt; Sent: Wednesday, October 17, 2018 7:40:52 AM To: wiki-research-l(a)lists.wikimedia.org Subject: Re: [Wiki-research-l] URL-addressable Predicate Calculus Interesting. Why would you have the prefix "https://machine.wikipedia.org/". This could be useful for many projects particular Wikidata, - not just wikipedia. Would "https://wikimachine.org/" or something be better? "mw" is used as prefix for https://www.mediawiki.org and mw2sparql https://www.mediawiki.org/wiki/MW2SPARQL "mw:P2<t0, t1, t2>" does not seem to be supported by SPARQL and < and > are not allowed in IRI_REF. Ordinary parentheses seems to be allowed. The rule is '<' ([^<>"{}|^`\]-[#x00-#x20])* '>' "mw:P1(arg0, arg1, arg2)" and "mw:P2<t0, t1, t2>" Why are there two kinds of input? Wouldn't one suffice. "mw:P1" Why are the identifiers prefixed with P? That collides with the Wikidata properties. "mw:P1(arg0, arg1, arg2)" This form does not allow for optional arguments. There is also the question of what should be returned. There should be room for error messaging. One idea I have not followed through is to make an empty SPARQL endpoint that would just provide SPARQL functions (extension functions). This way the function could possibly be used in a federated query. best regards Finn Årup Nielsen On 10/17/18 11:32 AM, Adam Sobieski wrote: _______________________________________________ Wiki-research-l mailing list Wiki-research-l(a)lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l

Am 18.10.2018 um 06:09 schrieb Adam Sobieski: Have you tried applying that technique to the Wikidata Query Service? <https://query.wikidata.org/> In particular, your approach seems quite similar to Linked Data Fragements as described on <http://linkeddatafragments.org/>. Wikidata's LDF interface can be found at https://query.wikidata.org/bigdata/ldf -- Daniel Kinzler Principal Software Engineer, Core Platform Wikimedia Foundation