Introduction
The goal of PROV is to enable interoperable interchange of provenance on the Web. We take two approaches to documenting the implementation and interoperability of PROV.
- For the data model [[PROV-DM]] and the two working group defined serializations, we document the existence of multiple implementations that support each of the constructs within the data model and that there exists at least two implementations that are reported to exchange these constructs.
- We document that the PROV-Constraints specification is implementable by documenting the existence of at least two implementations that report to support all the defined constraints. To evaluate the coverage of implementations, the PROV-Constraints Test Cases are used as a point of reference. There are 280 test cases in total, which map to the the constraints defined by the document.
PROV is useful not only for applications/programs but also for exposing provenance within datasets and as a foundation for other vocabularies. We also document that usage as well.
Implementation evidence was gathered using four surveys.
This report summarizes the results of these surveys.
Meeting the Exit Criteria
At the start of the Candidate Recommendation phase, the Working Group defined a series of exit criteria. These exit criteria can be summarized as for each feature defined by PROV there are at least two implementations that support the feature and that there exists one interoperability pair that can exchange that feature. Section 3.1 shows that three implementations both produce and consume all constructs defined in PROV-DM. PROV-O is implemented by over 40 implementations and PROV-N is implemented by 12 implementations.
In terms of implementation pairs, Section 4 enumerates which pairs of implementations report exchanging provenance. Here, we meet the exit criteria in that each feature is exchanged by at least two implementations.
Finally, three validators have implemented all of the constraints defined in PROV-Constraints passing the requisite test cases, thus, passing the exit criteria. The working group recognizes that implementing the PROV-Constraints document requires substantial effort. It is nice to see that three radically different approaches were chosen to implement this specification: SPARQL, Java, Prolog, which speaks to the implementability of this specification.
For a systematic enumeration of how the exit criteria were met, please see http://www.w3.org/2011/prov/wiki/MeetingProvCRExitCriteria
Acknowledgements
We would like to thank the following who reported their PROV implementations to us: Ali Mufajjul, Amir Sezavar Keshavarz, Ashley Smith, Chester Chen, Chris Baillie, Clément Caron, Daniel Garijo, Danius Michaelides, David Corsar, Edoardo Pignotti, Eric Stephan, Hook Hua, Irene Celino, Jacco van Ossenbruggen, James Cheney, James McCusker, Jens Lehmann, Jun Zhao, Kerry Taylor, Khalid Belhajjame, Landong Zuo, Luc Moreau, Luis M. Vilches-Blázquez, Michael Jewell, Mohamed Morsey, Olaf Hartig, Palma Raul, Paolo Missier, Paul Groth, Peer Brauer, Peter Slaughter, Reza B'Far, Rinke Hoekstra, Sara Magliacane, Sarven Capadisli, Satya Sahoo, Simon Miles, Spyros Kotoulas, Stephan Zednik, Stian Soiland-Reyes, Timothy Lebo, Tom De Nies, Trung Dong Huynh, Victor Rodriguez, Yanfeng Shu.
),
produced (
),
or both consumed and produced (
)
by a particular implementation.