Querying Linked Data on the Web,
for real this time

Problems

• Privacy issues

  Not all data is public (e.g. Solid), GDPR, ...

• SPARQL endpoints don't scale for many clients

  Public SPARQL endpoints down for 1.5+ days each month (95% availability).

• SPARQL endpoints don't scale for many sources

  Federation algorithms only work in the order of 10 sources.

• We have many small sources

  Websites use schema.org (36% in 2014), Solid data pods are popping up, ...

(Possible) Solutions

• Bring querying to the client

  Client-side engines authenticate themselves to sources.

  Better scaling, as clients do most of the work themselves.

• Link-traversal based querying (because clients can not index everything)

  Detect query-relevant sources on-the-fly by incorporating live crawling into the querying process.

• Use SPARQL endpoints as an optimization

  Clients can use this as an external index over certain sources.

(More) Intelligent Clients are needed
to query Linked Data on the Web