Requirements for Unique Identifiers      
  
1.  a.  A single framework should be used to identify anything a VO  
	application can refer to including resources, projects
	(mission/telescope), data collections, data objects, and
	services. 
  
    b.  The framework should support a notions of local and global  
	identifiers.  A local identifier is considered unique only  
	within a certain context.  A global identifier is considered  
	unique across all VO contexts.  It should be possible to  
	readily distinguish between a local and global identifier.    
  
2.  It should be easy to compare two identifiers to determine if the  
    objects are the same.  This requirement does not specify exactly  
    what constitutes sameness.	
  
3.  Users should be able to effectively use identifiers to avoid  
    retrieving multiple copies of the same dataset.    
  
4.  It should be possible to identify different versions of the same  
    data object in different formats.  (This might be done with a  
    single identifier or with an identifier and a MIME type.)  
  
5.  a. DROPPED: It should be easy to identify the resource with primary	 
       responsibility for a data object based on its identifier.   
  
    b. It should be easy to identify the resource having issued the  
       identifier.    
  
    c. It should be possible to use an ID to access a unique  
       description of the item it identifies; however, it is not  
       required that such a description be guaranteed to exist.    
  
6.  DROPPED: It should be easy to identify the collection to which a data  
    object belongs based on its identifier.  
  
7.  It should possible for data mirrors to reuse collection and data  
    object identifiers.	 
  
8.  DROPPED: It should be possible for an identifier to indicate that  
    an object or collection is derived from another object or collection.    
  
9.  The framework should maximize the freedom of data providers to  
    choose identifiers for resources and collections under their  
    control.  	
  
    a. The framework should minimize the restrictions on the  
       form of the identifier to that which meets the requirements.  
      
    b. There should be no single name granting authority; rather,  
       providers are merely required to choose identifiers that unique at  
       the time that they are registered.                                    


On Versions and Sameness:

The definition of sameness must recognize that multiple versions of
data will have several forms:

  * Byte-for-byte copies
  * Semantic copies of the same format.  e.g. two FITS files may contain
      the same image data but with slightly different headers or other 
      ancillary data
  * Semantic copies in different formats.  e.g. FITS vs. GIF.  Note:
      that the information content may be different in the different
      formats.  
  * Versions derived from the same source with different
      algorithms/parameters.  This includes the special case of
      cutouts.  
  * One version derived from another version.  

A definition of sameness should indicated which of these cases are
covered.  Note that the identifier framework need not necesarily
encode all of these relationships.  An analysis of ancillary
information (metadata) may be required to determine this.