Bag-semantics formalization for automatic federated SPARQL source selection

Establish a bag-semantics extension of FedQPL, or another formal definition of federated SPARQL queries under automatic source selection that reconciles FedQPL’s set-based execution with SPARQL’s bag semantics, to support reproducible federated evaluation.

Background

FedQPL is presented as a reference formalism for query plans generated by automatic source selection in federated SPARQL systems. However, FedQPL defines query answers and operators over sets of solution mappings, while SPARQL evaluates queries under bag semantics, in which duplicate solution mappings may be significant.

The paper identifies a concrete discrepancy for query S7: the same triple is present in two federation members, yielding one solution under set-union semantics but two solutions under the observed SPARQL execution. Because FedQPL provides no SPARQL union operator that bridges these semantic models, the expected behavior of real federated SPARQL engines under automatic source selection cannot be formally checked, undermining reproducibility.

References

FedQPL, the reference formalism for query plans that automatic source selection produces, requires the answer to a query over a federation to be the one obtained by evaluating it over the set union of the data of all federation membersDef.~3, that is, over the federation seen as a single knowledge graph, in which the triple asserted by both sources occurs once. Its operators are correspondingly defined over sets of solution mappingsDef.~6, whereas SPARQL operators evaluate under bag semantics, and the formalism proposes no union operator for SPARQL that would bridge the two, so how to reconcile them is not settled. A bag semantics for FedQPL is left to future work, and no published extension develops one.

— Strengthening LargeRDFBench for Interoperable Federated SPARQL Evaluation  (2608.30465 - Tam et al., 31 Aug 2026) in Section 3, Experiment; reiterated in Section 4, Conclusion