On 2 May 2008, at 15:48, Jewel Ward wrote:
On May 2, 2008, at 5:24 AM, Phil Cross wrote:
But that doesn't disguise the large amount of resources that are required to normalize IR metadata. You could argue that isn't a fault with the OAI-PMH protocol, though, but is due to the simple fact that there is no standard format for exporting metadata from repositories - even though they all have to use Simple DC, there are quite a few ways in which that is used. And I think that problem will be there no matter what transport protocol is used to exchange the metadata.
as someone who shall remain nameless said to me in 2003, "thanks for proving the entire [OAI-PMH] framework is doomed to fail" because of the lack of use of metadata, period, or the lack of quality metadata. (said in reference to the results of my master's paper results, to toot my own horn).
"Doom"? "Fail"? "Proof"? That's fighting talk!
The OAI framework is an interoperability framework that separates out the roles of data providers and value-added service providers, and does so by providing a metadata harvesting framework. It was very forward looking for its time (driven by the need to escape from uninterpretable information silos), and I would still argue passionately for its relevance today.
However, it was arguably shaped by constraints that became irrelevant, such as the perceived need for a controllable batch protocol to manage load on under-resourced servers.
A key question raised here is whether the obsession with metadata is part of the forward thinking, or part of the outdated thinking? I think the former - there has to be a communication about the holdings of a repository to the service providers, so that they can provide the service.
Historically speaking the major service that was anticipated was cross-repository search, and that was completely blown away by the Googl-ification of the academic world. Now Google search has worked very well for all of us, but it has not needed the OAI framework because it uses a less sophisticated model of information communication - if you want to use an item, download it! But it has its costs - a huge data infrastructure with lots of air-conditioning backing up constant crawler downloads.
Is the metadata framework dead then? I believe not, although perhaps a batch harvest is too unwieldy for modern web-programming paradigms. In essence, the OAI framework is very similar in purpose to the W3C's Linked Data initiative, and I think that we shall see some reconciliation between them in the short term.
the work that Naomi Dushay and Diane Hillmann did in examining the NDL metadata, and that Tim Cole and Sarah Shreeves, et al, have done with the IMLS DCC project (http://imlsdcc.grainger.uiuc.edu/about.asp) has shown that we are in no better a position now in terms of shareable quality metadata than we were at the start of the OAI-PMH, when it was UPS, oh, 9 years ago? and, unlike 9 years ago, folks **know** they need to provide quality metadata, and provide metadata, period. 'm not just referring to the DCMES, but **all** of the metadata formats.
Does this prove that metadata or PMH is dead? No way! I think rather it shows how long it takes to go from new interoperability protocol, through the development of usable and useful embodying implementations (the repository platforms), through to mainstream adoption within a library context, through to embedding within the institutional community working practices. We've been developing EPrints ever since the original OAI/UPS meeting and it's only NOW that we've got around to starting to provide metadata QA management support, because there are so many other hurdles to jump first!
In the UK there has been a recent driver for metadata quality for institutions that have used repositories in the current national Research Assessment Exercise. It will be very interesting to examine metadata quality in the aftermath of this externally-imposed QA process!
By the way, I can't see a Cole and Shreeves presentation in the URL you cite, but the Sarah's presentation which discusses OAI & Metadata problems (
Search Interoperability,
OAI, and Metadata) was given in 2005, which is the same year that UIUC started their repository. So I don't agree that it represents 9 years of accumulated experience.
if there is or was a problem with the OAI-PMH, i think it was more the assumption that there would be quality, shareable metadata on which to build services, more so than a problem with the framework itself as part of the Web architecture or the implementation of it.
i'm skeptical that orgs that cannot implement the OAI-PMH or found it "difficult" to implement and manage will be able to handle something as complicated as FRBR, et al, no matter how wonderful it is/they are, because of the technical support/resources/knowledge required.
Organisations do not (on the whole) implement PMH, any more than they implement SMTP. They install an EPrints/DSpace/Fedora server like they do an Exchange server. Their users might provide unsatisfactory metadata, but then they might send unsatisfactory email too :-)