Data provenance is de ned as information about entities, activities and people producing or modifying a piece of data. On the Web, the interchange of standardized provenance of (linked) data is an essential step towards establishing trust [2]. One mechanism to track (part of) the provenance of data, is through the use of version control systems (VCS), such as Git. These systems are widely used to facilitate collaboration primarily for both code and data. Here, we describe a system to expose the provenance stored in VCS in a new standard Web-native format: W3C PROV [4]. This enables the easy publication of VCS provenance on the Web and subsequent integration with other systems that make use of PROV. The system is exposed as a RESTful Web service, which allows integration into user-friendly tools, such as browser plugins.
Version control systems (VCS) have a long history in computing. The rst such
system was the Source Code Control System, developed in 1972 [
Versioning of data is an aspect of provenance: information about entities,
activities and people producing or modifying a piece of data. Provenance is critical
in contexts ranging from scienti c reproducibility to journalism. Furthermore, a
number of sub-disciplines of computer science, including databases, distributed
systems and the Web have been addressing issues related to provenance [
Thus, the aim of the system described in this paper is to enable the provenance within VCS to be exposed in a Web-native and interoperable format. This provenance can then be consumed by other PROV enabled tools. Indeed, given that the software used to create many Linked Data sets is available through public version control systems6, we believe that the tool can be used to enrich the provenance of many of these datasets.
The rest of this paper is organized as follows. We brie y discuss related work and present a mapping from the Git version control system to PROV. This is followed by a description of the implementation and demonstration of our system. In particular, our demo illustrates how the resulting information can be consumed by other PROV enabled systems. The demonstration (live and video) is available at the following URI: http://git2prov.org. 2
This paper is part of a greater e ort to create more interoperability among
di erent platforms that track provenance information, by mapping them to a
standard interlingua such as [
Currently, the most commonly used VCS is Git, an example of a distributed
version control system. Due to brevity we will omit a detailed explanation of
Git, and refer to [
Our mapping, shown in Fig. 1, was created by identifying whether the data could
represent information from one or more broad classes of provenance information.
The three classes we used are identi ed below. For each class, we describe how
provenance can be expressed using concepts from the PROV Data Model [
{ Attribution - attribution information expressed as the prov:Agent that created the Entity using the prov:wasAttributedTo and prov:wasAssociatedWith relations, modeling the two potentially distinct roles of a author and a committer.
These classes re ect the three use-case perspectives on provenance identi ed by
the W3C Provenance Primer [
Because we want our conversion tool to be as exible as possible, we chose to build a RESTful git2prov Web Service for this purpose, using the Node.js7 framework. The service is available at the following URI: http://git2prov. org/git2prov?gituri=<your_git_uri>.
The only required input for this service is a URL giturl that refers to a git repository. In this rst proof-of-concept implementation, only openly accessible repositories are supported. However, adding support for secure repositories is part of our future work. In addition to giturl, the service accepts a number of optional parameters, with the default value in bold: serialization (possible values: [PROV-N, PROV-O, PROV-JSON, PROV
XML]) This parameter is used to specify the desired PROV serialization. shortHashes (possible values: [false, true]) This parameter forces the service to use the short commit hash in the exported provenance, to increase readability for human users 7 http://nodejs.org ignore (possible values: a provenance relation) This parameter is used to lter the speci ed relation from the converted provenance.
Note that each provenance document generated by the git2prov service includes a link to the complete document (without any restricting parameters).
Upon receiving a request, the service clones the git repository to a temporary location, and performs a git log command on it. The output of this log is then mapped to PROV as described in Sect. 3, and written to the HTTP response in the requested serialization. To demonstrate our service, we wrote a small web application, available at http://git2prov.org. A video is also available at http://vimeo.com/70980809. 5
We believe that systems such as Git2PROV have the potential to become an important enabler of the widespread interchange of standardized provenance. With our proof-of-concept implementation, we have shown that it is certainly feasible to build a lightweight RESTful Web service to convert versioning systems into PROV. In future work, we aim to improve our work by including more semantic annotations in combination with the provenance to allow further reasoning over it, with the prospect of deriving trust assessments.
Acknowledgements These research activities were funded by Ghent University, iMinds, the IWT Flanders, the FWO-Flanders, and the European Union, as well as the Data2Semantics project in the Dutch national program COMMIT.