Linked Data are modeled following the RDF data model: every concept can be mapped into this information model that is not bound to any specific type of data. Spatial information or alphanumeric can be expressed according this simple model where each unit of data is always a triplet of
(subject, predicate, object)
that resembles the human way of thinking and of representing the external world. "The water is hot" or "the lion eats the zebra" are example of concepts that can be translated into few RDF triples: (water, is a, chemical substance), (water, has temperature, hot) and (lion, is a, animal), (zebra, is a, animal) and (lion, to eat, zebra). The RDF triples can be stored in special databases called "triple stores". This software component not only allows the storage of the RDF triples, but offers a standard query language: SPARQL.
SPARQL is pronounced like sparkle and is a recursive acronym meaning Sparql Protocol And Rdf Query Language. Although its syntax is quite simple and resembles the more known SQL, people usually find it rather unnatural; a SPARQL query simply asks the system to extract all the triples that match the triple templates written in the WHERE clause.
For instance, the query:
?subject rdf:type foaf:person.
?subject foaf:name ?name.
?subject foaf:mbox mailto:firstname.lastname@example.org.
extracts all the triplets (actually just the subject's name) that have the email address equal to mailto:email@example.com. All those three triplets' patterns in the WHERE clause must be tried against all the triplets in the store and the ones that pass the filter are presented as the result.
Unlike a SQL one, a SPARQL endpoint is always accessible via HTTP protocol and the query can be url-encoded and passed as a parameter to the service.