[OAI-implementers] Query Regarding OAI Identifier
simeon.warner at cornell.edu
Thu Apr 9 16:43:55 EDT 2009
Pope Alison wrote:
> We are working on a project that will take our current eprints collection
> from a dedicated eprints repository platform to a new repository platform
> where eprints will be one of many collections. I?m trying to understand
> what the OAI implications of this will be and what might be the impact of
> harvesters and other service providers. One of the things is with this
> broadening of the repository is it will use a different domain name as
> currently designed. I note however that this goes against the OAi
> Identifier principles as we would want our namespace identifier to change
> also. This is currently defined as:
> We might in future want it to be:
> The OAI data provider registration guidelines
> (http://www.openarchives.org/data/registerasprovider.html) states a
> repository can only have one namespace identifier and a repository may not
> change its namespace identifier. Does this mean:
> a) It?s not possible to change the OAI identifier as specified above?
This is not enforced, it is just bad practice. Any services consuming the
metadata may get confused...
> b) We could change the OAI identifier above but we need to remove all
> registrations for our old repository and register this as a new repository
> rather than change the repository information
Should really also tell all services consuming your metadata to delete it
too, and then ask them to start again with new setup.
> c) Continue to use our existing eprints.rhul.ac.uk domain in the OAI
> identifier configuration even though the repository URL is now
> digirep.rhul.ac.uk because the identifier only relates to the OAI
> request/response harvesting.
This would be to best option to persist the identifiers and data. However,
if you change location of your server you'd still need to let consuming
services know that.
I suppose the bottom line is that OAI-PMH does not have mechanisms to deal
with this sort of change. One has either to resort to web-server tricks to
make old URIs work, or else talk to consuming services to push
(Putting a technology name in the URI has just the same problems that all
those silly URLs out there with .jsp, .asp, .cfm, .php extensions have.
When the technology changes you either end up with confusing URLs or you
have to break or redirect old ones...)
> Has anyone done a similar repository migration and considered these kinds
> of issues or has any suggestions on what best practice would be in this
> sort of scenario.
> May thanks,
> *Alison Pope*
> Business Analyst
> IT Business Development
> Computer Centre
> Royal Holloway, University of London
> TW20 0EX
> alison.pope at rhul.ac.uk <mailto:alison.pope at rhul.ac.uk>
> 01784 414925
More information about the OAI-implementers