The Mellon Foundation prefers open source software because they prefer not to be locked into any particular solution.
They have provided an exception in this case (based on justifications provided in RS Technical Offer, in particular section 3.1.1) but require an answer to this question:
- “Re. the RDF store: we would like some additional assurance that the Ontotext RDF store, though proprietary, will not pose unnecessary difficulties should the database choose (or need) to move to an open source store in the future. In essence we know it is possible to move, but how possible? Are there dependencies that need to be acknowledged?”
The RDF repository is used mostly in an OWLIM-independent way, based on the following aspects:
- Access the repository through the most popular open source semantic API (Sesame). (OWLIM also supports the other popular API, Jena)
- Standards compliant features of RDF, named graphs, SPARQL
- Repository updates will be implemented using SPARQL 1.1
OWLIM-specific dependencies are the following:
- Full-text-search indexing within the repository, using OWLIM-specific predicates (uses the open source Lucene FTS engine). The Jena TDB open source repository has similar capabilities through LARQ
- Implementing CIDOC CRM Fundamental Relations (used for Semantic Search).
- The few FRs currently implemented use OWL2 RL, but it turns out that OWL2 is not expressive enough to define new properties. This answer suggests it's impossible to define a property by conjunction in OWL2, even OWL2 Full. This paper "shows that adding role conjunctions to DLs causes a jump in the computational complexity of the standard reasoning tasks.. due to a subtle interaction between inverse roles, role hierarchies, and role transitivity with role conjunctions"
- Therefore when we face more complex FRs in further iterations, it is likely we'll have to reimplement them using OWLIM Rules
- Incremental assert/retract. OWLIM is one of the few repositories that offer incremental assert/retract of triples (i.e. inserting, updating and deleting single facts), while supporting the required inferences (CIDOC CRM subclass and subproperty, and Fundamental Relations).
- This can be simulated in other repositories using offline (nightly) reindexing, the inconvenience being that newly entered data cannot be searched, and expected inference conclusions will not be present, before the nightly update
- Note: the OWLIM FTS index is not updated incrementally, though it can be updated on every commit, so it faces as similar issue but to a smaller extent