Dereferencing URLs in RDF Languages

[AdamSobieski]

Introduction

Drawing inspiration from pointers in computer-programming languages like C and C++, proposed for RDF languages is that the * symbol could be used to dereference URLs into: IRIs, literals, triples, quads, graphs, and datasets.

Single IRIs

Let us consider the following syntax:

*<http://www.data.org/example/1.uris> a ex:Example .

The * symbol indicates to dereference the URL. The URL http://www.data.org/example/1.uris, for example, resolves to:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: text/uri-list

http://www.namespace.org/#fragment

This would result in:

<http://www.namespace.org/#fragment> a ex:Example .

Similarly, a URL could be dereferenced in the predicate or object position of a triple. That is:

ex:example a *<http://www.data.org/example/1.uris> .

would result in:

ex:example a <http://www.namespace.org/#fragment> .

Multiple IRIs

What if a resource of type text/uri-list provided multiple IRIs?

Multiple triples would be generated as a result of the multiple URIs/IRIs occurring in a text/uri-list resource.

So, for the following syntax:

*<http://www.data.org/sets/dogs.uris> a ex:Dog .

and with the URL http://www.data.org/sets/dogs.uris resolving to:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: text/uri-list

http://www.namespace.org/#fido
http://www.namespace.org/#rover
http://www.namespace.org/#spot

the following triples would result:

<http://www.namespace.org/#fido> a ex:Dog .
<http://www.namespace.org/#rover> a ex:Dog .
<http://www.namespace.org/#spot> a ex:Dog .

Similarly, a URL could be dereferenced in the predicate or object position of a triple. That is:

ex:alice ex:hasPet *<http://www.data.org/sets/dogs.uris> .

would result in:

ex:alice ex:hasPet <http://www.namespace.org/#fido> ,
                   <http://www.namespace.org/#rover> ,
                   <http://www.namespace.org/#spot> .

Literals

With a new MIME type, application/literal, literals could be represented as URL-addressable resources and processed into RDF-inline literals. This would work for literals of all sizes, from small text strings to very large data blobs.

ex:thing rdf:hasData *<http://www.data.org/example/3.lit> .

So, for the URL http://www.data.org/example/3.lit, a Web server might serve:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: application/literal

Hello world.

which would result in:

ex:thing rdf:hasData "Hello world." .

A Web server might, alternatively, serve:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: application/literal
Content-Language: en

Hello world.

which would result in:

ex:thing rdf:hasData "Hello world."@en .

A Web server might, alternatively, serve:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: application/literal; datatype="http://www.w3.org/2001/XMLSchema%23integer"

42

which would result in:

ex:thing rdf:hasData "42"^^xsd:integer

By encoding literals' datatypes as MIME-type parameters, HTTP content negotiation could include datatypes, enabling content negotiation with respect to numerical precision and units of measure, for examples.

Other possibilities include: multiple MIME types, e.g., application/literal+json or application/literal+cbor, could be utilized to encode, alongside data, the other components of RDF literals; and new HTTP response headers could encode the other components of RDF literals.

Single Triples

Considering graphs with one triple, we could make use of RDF formats like text/turtle.

The following syntax:

<< *<http://www.data.org/example/4.ttl> >> ex:attributedTo ex:bob .

with the URL http://www.data.org/example/4.ttl resolving to:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: text/turtle

@prefix ex: <http://example.org/stuff/1.0/> .

<http://www.namespace.org/#fido> a ex:Dog .

would result in:

<< <http://www.namespace.org/#fido> a ex:Dog >> ex:attributedTo ex:bob .

Multiple Triples

What if the graph returned by a dereferenced URL contained multiple triples?

The following syntax:

<< *<http://www.data.org/example/5.ttl> >> ex:attributedTo ex:bob .

with the URL http://www.data.org/example/5.ttl resolving to:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: text/turtle

@prefix ex: <http://example.org/stuff/1.0/> .

<http://www.namespace.org/#fido> a ex:Dog .
<http://www.namespace.org/#rover> a ex:Dog .
<http://www.namespace.org/#spot> a ex:Dog .

would result in:

<< <http://www.namespace.org/#fido> a ex:Dog >> ex:attributedTo ex:bob .
<< <http://www.namespace.org/#rover> a ex:Dog >> ex:attributedTo ex:bob .
<< <http://www.namespace.org/#spot> a ex:Dog >> ex:attributedTo ex:bob .

Blank Nodes

What if developers could dereference URLs into blank-node contexts?

The following syntax:

ex:example ex:currentWeather [ *<http://www.data.org/example/weather.ttl#this> ] .

with the URL http://www.data.org/example/weather.ttl resolving to:

HTTP/1.1 200 OK
Date: Wed, 19 August 2026 12:28:53 GMT
Content-Type: text/turtle

@prefix ex: <http://example.org/stuff/1.0/> .

<#this> a ex:WeatherReport ;
        ex:hasTemperature "73"^^ex:fahrenheit .

would result in:

ex:example ex:currentWeather [ a ex:WeatherReport ;
                               ex:hasTemperature "73"^^ex:fahrenheit ] .

The node in the dereferenced URL's resource indicated using a fragment identifier would be the one merged into the containing blank node.

Collections

What about lists?

ex:example ex:predicate ( *<http://www.data.org/example/7.uris> ) .

Ideally, the one or more IRIs in a text/uri-list resource would become the elements of the list.

TriG

In TriG, using the * syntax, in addition to all of the above possibilities, one could also:

ex:Graph1 { *<http://www.data.org/example/8.ttl> }

Dynamic Content

Web servers can generate dynamic content of types text/uri-list, application/literal, and text/turtle.

*<http://www.data.org/example/9.php?arg=value> a ex:Example .

ex:example ex:hasTimestamp *<http://www.data.org/example/current-time.php> .

ex:example ex:currentWeather [ *<http://www.data.org/example/current-weather.php#this> ] .

Multipart MIME Messages, Archive Files, and Packages

In addition to absolute and relative URLs using the HTTP URL scheme, the Web, one could use the cid: URL scheme for multipart MIME messages or relative URLs for resources within archive files or packages.

ex:thing ex:hasData *<cid:part2> .

<< *<cid:part3> >> ex:attributedTo ex:person .

ex:thing ex:hasData *</path/to/resource.lit> .

Conclusion

What do you think about these ideas? Thank you.