organization/location-independent identifiers

Ray Plante rplante at poplar.ncsa.uiuc.edu
Thu Sep 11 07:58:11 PDT 2003


Hi, 

As Arnold outlined, there is an important driver for persistant, 
location-independent identifiers coming from the journals, and I think it 
is important we try to get some compatibility between the IVOA and ADEC.  

To be a bit more precise, the IVOA Identifier WD describes 
organization-dependent identifiers.  That is, as long as a single 
organization continues to curate a resource, it can freely move its 
location or mirror it at several locations (perhaps hiding these details 
behind a single interface) without changing the resource's IVOA 
identifier.  

However, if curation of a resource is transfered to another organization, 
or if another organization decides to curate the resource, it is 
considered a new instance of the resource; thus, it will have a different 
identifier and metadata description. 

Given that we need to distinguish between different instances of a 
resource, I believe we need to have a second type identifier to handle 
location independence.  There is precedence in the internet world for this 
in the distinction between URIs and URNs.  The latter is what we need; 
however, it would help if they were related in some way. 

Here's my strawman, with many details missing:

  1.  we create metadata that allows resource descriptions to say:
       *  "I am a mirror of this other resource"
       *  "The following resources are recognized mirrors of me"

  2.  When a new standard service is defined, it should include an 
      explanation of what it means for two instances of that service to be 
      mirrors.  

  3.  Resources use the metadata from #1 to claim a mirror relationship. 

  4.  #3 is not sufficient, because when we have mirrors of mirrors, it is 
      difficult to traceback through all the relationships.  All mirrors 
      should share a common URN; when someone mirrors another resource, 
      they adopt that URN.

  5.  Journals can use the URN to cite datasets in articles.

cheers,
Ray







More information about the registry mailing list