[OAI-implementers] OAI identifier resolver

Young,Jeff jyoung@oclc.org
Mon, 20 Oct 2003 12:24:38 -0400


Just so we're clear, this service does indeed only work in conjunction with
a single "centralized" OAI-harvestable registry. That is by design. My hope
is that Tom Habing's registry will eventually become harvestable, so I can
point at it instead of my dummy registry. My service doesn't really care how
the repositoryIdentifiers are assigned in the Registry (hashed, sequential
numbers, or manually assigned), as long as it's embedded in the identifier
for each registered repository in some structured way. 

Keep in mind, though, once a repositoryIdentifier is assigned, my service
will fail to resolve old URLs if it is changed from something ugly to
something meaningful (unless the old identifier is retained in the registry
as well).

If repositories don't want to register with Tom, it would be possible to set
up other private registry/resolver combinations for particular communities.
Perhaps Tom would consider distributing his Registry software as open-source
to facilitate this need.

If a repository doesn't want to register anywhere, then it's still possible
to create a set of PURL partial redirects that look like the ones you
proposed earlier.

Jeff

>  -----Original Message-----
>  From: Xiaoming Liu [mailto:liu_x@cs.odu.edu]
>  Sent: Monday, October 20, 2003 12:03 PM
>  To: Young,Jeff
>  Cc: 'Xiaoming Liu'; OAI-implementers (E-mail)
>  Subject: RE: [OAI-implementers] OAI identifier resolver
>  
>  
>  On Mon, 20 Oct 2003, Young,Jeff wrote:
>  >
>  > - My hope is that these URLs will be as natural-looking as 
>  possible, which
>  > is why I'm advocating the assignment of meaningful 
>  repositoryIdentifiers
>  > during the registration process, even for repositories 
>  that don't use the
>  > oai-identifier schema.
>  
>  I think we all agree it's useful to uniquely identify a 
>  repository and all
>  its records in an URL-friendly way. Thus different service 
>  providers and
>  data providers can easily interoperate.
>  
>  I just personally feel it's probably easier to agree on an 
>  algorithm than
>  a centralized registration mechanism. MD5 generated 
>  fingerprint is probably
>  too long, but other hashing algorithms (like CRC32) can generate much
>  shorter signature.
>  
>  xiaoming
>  
>  
>  >
>  > Jeff
>  >
>  > >  -----Original Message-----
>  > >  From: Xiaoming Liu [mailto:liu_x@lanl.gov]
>  > >  Sent: Wednesday, October 15, 2003 1:47 PM
>  > >  To: Young,Jeff
>  > >  Cc: OAI-implementers (E-mail)
>  > >  Subject: Re: [OAI-implementers] OAI identifier resolver
>  > >
>  > >  Two comments:
>  > >
>  > >  -- If URL takes the format
>  > >  http://purl.org/oai/<metadata-format>/<oai-identifier>, it
>  > >  will make it
>  > >  much easier to parse the special format identifier like
>  > >  hdl:1721.1/659,
>  > >  the program will know whole part after "metadata-format" is
>  > >  oai-identifier.
>  > >
>  > >  -- I wonder if it's useful to assign repository ID by hashing
>  > >  the baseURL, this will make it possible to develop
>  > >  applications without
>  > >  prior agreement on unique ID of repositories.
>  > >
>  > >  For example, the baseURL of DSpace 
(http://hpds1.mit.edu/oai/) can be
> >  hashed to 2D0DA5560CAFBA3912A522345ECA4271 by MD5 algorithm.
> >  If both POI
> >  and UIUC resolvers use this algorithm, the same repository ID can be
> >  generated without prior agreement.
> >
> >  xiaoming
> _______________________________________________
> OAI-implementers mailing list
> List information, archives, preferences and to unsubscribe:
> http://oaisrv.nsdl.cornell.edu/mailman/listinfo/oai-implementers
>