LinkedDataContentNegotation

A proxy setup for Linked Data Content Negotation

This Apache 2.4 proxy setup implements the Flemish Government URI strategy based on id / doc using 303. This implementation will be the reference implementation of this strategy so that others within the Flemish Government can quicky exploit this. Additionally the default pages are also according to the Flemish Government webstyle rules.

Despite the tailoring towards the Flemish Government look&feel and some URI decisions, this forms still a template implementation that can be applied by any Linked Data publication platform. Therefore the design choices are documented here so that others organisations which like to deploy a Linked Data publishing platform can move forward quicker.

Deployment structure

The service based upon the official Docker Container for Apache2 webserver.

The service is organized as follows:

@build-time

• /config : the configuration files
• /scripts : the cgi-scripts
• /www : the document root

@runtime

• /logs : the logs

Architectural embedding

The service connects to other services, which should be declared as accessible hosts:

• ldsb-service : Linked Data Subject Block service
• subjectpages-service : HTML subject pages service
• sparql-endpoint-service : SPARQL endpoint

The service URLs are given by the following environment variables

Environment variable	default value
ENV_LDSB_SERVICE_URL	http://ldsb-service:81
ENV_SUBJECTPAGES_SERVICE_URL	http://subjectpages-service
ENV_SPARQL_ENDPOINT_SERVICE_URL	http://sparql-endpoint-service:8890/sparql

Configuration

Environment variable	default value
ENV_URI_DOMAIN	data.vlaanderen.be

Execution

A typical start in production setting will be

docker run -d -p 80:80 
    --add-host sparql-endpoint-service:<ip/hostname>
    --add-host ldsb-service:<ip/hostname>
    --add-host subjectpages-service:<ip/hostname> 
    -e ENV_LDSB_SERVICE_URL=http://ldsb-service
    --name=cn 
    -v /persistentstorage/logs/LinkedDataContentNegotation/:/logs
    bertvannuffelen/linkeddatacontentnegotation

Note the public exposure of the logs on persistent storage. It is a good practice to ensure that the logs are stored on a safe location. If the service has to change, the logs are kept for future problem resolution.

Design decisions

For instance, on places where a 300 Multiple Choice repons usually is suggested, this implementation will respond with a 404 providing a custom message indicating the supported accept headers. This complies to the current standardization levels and best practices for implementing the HTTP responses. Namely the HTTP 300 respons has no standard respons body which can detail the allowed/supported options by the server. Therefore the handling of the HTTP 300 respons cannot be automated, and it it better to force the consumer to take explicit action in its client code.

sparql endpoint

This proxy activates a sparql-endpoint at the url $ENV_URI_DOMAIN/sparql. This sparql interface has a web interface and a machine readable interface. Via content negotation the machine readable interface is reachable. The Accept header is one of the following values defined according to W3C:

ACCEPT header	description format
application/sparql-results+json	[https://www.w3.org/TR/sparql11-results-json/]
application/sparql-results+xml	[https://www.w3.org/TR/rdf-sparql-XMLres/]
text/csv	[https://www.w3.org/TR/sparql11-results-csv-tsv/]
text/tab-separated-values	[https://www.w3.org/TR/sparql11-results-csv-tsv/]

content negotation

This proxy activates content negotation. The table below describes the implemented accept headers.

document format	corresponding accept-header	corresponding extension
HTML	text/html	.html
RDF	application/rdf+xml	.rdf
TURTLE	text/turtle	.ttl
NTRIPLES	text/ntriples	.nt
jsonld	application/ld+json	.jsonld
json	application/json	.json

Notes

a) https://pwc-technology-be.github.io/oslo/dienst/dienst.jsonld is used instead of the github version. This is a manual update for now.