The most important requirement for RIOXX is to provide a way to collect basic information about how research outputs (especially papers) have been funded. That is to say, we need to collect two types of information for a given paper:
RIOXX has at its heart a simple application profile of Dublin Core, the better to fit naturally with existing technical infrastructure and to be easily implemented. Following from this, it seems that there are the following possibilities for representing funder and projectID in a single record:
This is the currently proposed approach. It has the virtue of introducing a very low barrier to data collection (the common repository interfaces support this) and provides for the primary use-case which is the collection of funding information about research outputs. In data-modelling terms, this is not a satisfying approach because it is not possible to relate a funder to a project ID within any given record. However:
This is not a viable option as data transfer using XML does not reliably support the ordering of values within multi-value fields.
This is an obvious choice for an approach which does not depend on any extra external development (for example of a global unique ID system for projects/grants). However, this approach could be disruptive to the interfaces and workflows faced by repository managers, and we are aiming for minimal disruption to repository managers who are, after all, a key stake-holder. This option is still open for discussion and, indeed, we are discussing this with developers from the ePrints and DSpace communities.
This approach concatenates funder and projectID into one string, using an agreed delimiter character to separate them. This is the approach used in the OpenAIRE guidelines, and one which we have considered closely. We have decided to reject this approach because the projectIDs we will need to accommodate are unpredictable, and liable to contain whatever character we use to delimit this string. There are approaches to mitigating this issue - notably using encoding approaches such as URLEncoding - but this places an extra processing burden on suppliers and consumers with no guarantee that this will be implemented - this approach is therefore considered non-optimal and 'fragile'.
This is a more aspirational option - it cannot be implemented in the short-term because a system of globally uniques and persistent identifiers for grant-funded projects does not yet exist. However, in the medium-term this is feasible and is something we hope to explore in a later phase.
We are now working with experts from the ePrints and DSpace communities to figure out what is viable in the short term. Our default position is to implement option 1. If we can implement option 3 without damaging the chances of the overall solution being adopted by repository managers on the common platforms, then this would be an attractive solution for the short term. For the longer term, we will continue to consider approaches to option 5 or, better still, a combination of option 5 and option 1, where the funders are also listed separately to the projectID (and actionable and persistent identifier which points to metadata about the funder), simply to avoid the need for multiple round-trip queries to, for example, list records with a given funder.
Comments on all/any aspects of this are welcome!comments powered by Disqus