Thanks to everyone who participated in the Beyond OAI-PMH session this morning. There seems to be a number of places where others have posted their session summaries, but I thought it would be useful to include that here on the LOD-LAM site as well.
This is my brief (and somewhat tardy) account of our meeting based on the notes that I jotted down. Comments and responses are welcome and any misrepresentations are my own.
Two main themes:
1. We need to leverage existing OAI-PMH installation base for Linked Data, because after all it does fit within the basic requirements (three stars) of Linked Data goals.
- Promote adoption of tools like OAI2LOD server
- Build Sitemaps using existing OAI services
We should acknowledge current OAI-PMH for their existing contributions to Linked Open Data and emphasize that they already are participating in LOD through OAI-PMH. While there are some additional things we can do to make the metadata we share more LOD friendly, LOD is not a completely new idea. Additional documentation about how OAI-PMH has succeed and failed – and what lessons that holds for the future of LOD – would be welcome.
2. We don’t necessarily need an OAI-PMH 3.0
It would be better for the community to look towards broadly adopted web standards. Repositories need to provide what users want in multiple serializations, not limited to XML (let alone a specific XML schema).
Some suggestions for alternatives:
- Sitemaps
- Open Search
- Atom
While there wasn’t a strong sense that a new OAI standard was needed, there is a recognized need to provide the existing repositories some guidance about alternative approaches. Such guidance should be promoted by funders to help new and existing projects understand how they can contrbute to the Linked Data cloud. There was also a sense that some features of the current OAI protocal might be included in the development of web services:
- ability to acquire incremental sets (what’s changed, what’s new)
- an understanding of the “scope” of what’s provided (OAI sets/collections)
- a minimal set of shared properties (a stub) that is linked directly to richer representations
- some consensus around shared service models to make discovery and use easier
- ability to request sets based on supplied criteria (“search” not just pre-constructed sets)
Of course the devil is in the details, during the session we had several tangential conversations about the technical details of how to implement some of these alternatives that I haven’t fully captured here. To me this indicates that further discussion about these different options and how they might be shaped into a common framework is needed and would be valuable guidance for the community.
Additional Comment
There was also a suggestion that OAI-PMH may still be the best way to share large sets of “records” between partners. Rather than worry about making OAI-PMH more LOD friendly, LAMs may wish to focus their energy on providing other kinds of data as LOD (use cases?)