Saturday 24 November 2007

Getting It Out of your System

At the moment I am torn between the model of a repository as a theme park (there's all the rides for you to enjoy in one place) or the repository as a DVD lending library (there's all the films there in one place, but you take them away to enjoy them on your DVD player and your iPod and your laptop and at your friend's house).

I've thought hard about this with EPrints in mind - should it offer a rich and engaging user experience there, inside the repository, with as many built-in services as possible, or should it just let you take its contents and use them in as many external places as possible?

At Southampton we build repositories (EPrints) and we build OAI services (citation analysis, open access monitoring, preservation assistance), so we're constantly asking ourselves the question "where does this go? Inside a repository or in a service?" My natural inclination is to go for the external service model - it's global and interoperable rather than parochial and platform-specific.

Put in those terms the answer seems to be a no-brainer. But the problem is that while it is getting easier to get repositories funded and supported, it is really difficult to get services funded and supported. What is the natural home for an international service? Difficult to say! If it doesn't cater to a particular institution, region or country then who is going to put their hands up and host it? Or rather who is going to put their hands in their pocket and bankroll it?

But when it comes to an institutional repository there's a different story. It has a natural home (the institution) and with it a support infrastructure and a mechanism for applying for further support to achieve new developments/updates, all because it's serving a local need. So innovations and services may perhaps emerge in the local repository, rather than in a global service.

Tim O'Reilly made a recent criticism of this approach (It's The Data Stupid) in the context of social networking, arguing that it is more important to allow users to use their information in lots of third party services than it is to make it easy for developers to create lots of applications local to a particular site.

And he may be right - the information captured by a single repository is going to be a very very small part of "the global literature". What researcher would want to be locked up with only the work authored by him/her and his/her research group, however interesting the ride!

So on the one hand external services seem to be the proper solution and on the other hand local repositories seem to be the pragmatic solution. Like I said at the beginning, I'm torn. I think that EPrints had better back both approaches!


  1. This was a topic of conversation at the recent UKCoRR meeting. My inclination is also for the external service model. It's also worth noting that JISC are promoting the use of the e-framework ( which explicitly addresses the creation of applications as Services. JISC Repository start-up and enhancement projects are also required to create Service Usage Models to identify how the project might inter-operate as a service. With this in mind, it shouldn't be difficult to get funding and support for the creation of application services.

  2. You need both, recognizing that internal and external users have different needs. An external user won't care about document versioning; s/he just wants the latest greatest. An internal user cares a LOT. Mutatis mutandis, internal users won't care about RSS feeds, but external users who write mashups do.