Summary on ID trinity

[Arnold Rots]

Juggling names around is sometimes helpful.
Here is what I was suggesting in slightly different words.

A complete identifier consists of three sub-identifiers:

L:P:D

L a service resource or location
  Has validity in the service resource namespace
  It defines a location namespace - a data repository or a URL domain,
  if you like: MAST, IRSA, HEASARC, CDA, ADS, etc.

P a project or service
  Has validity in the namespace that contains all projects: missions,
  observatories, telescopes, services such as exposure time calculators
  It defines a project namespace such as CXO, HST, RXTE, VLA, etc.

D a dataset or object
  Has validity within a particular project namespace and points to a
  dataset, observation, file, piece of information, concept, etc.


Each project needs to register its project namespace identifier.
It also means that each service resource location has to register its
identifier, as well as the list of project identifiers that it serves.
Within a project, the originator is free to design and specify its own
dataset identifiers.  It may be useful to disallow certain symbols
such as :;%&? and suggest that / be used as separator.

One should be able to refer to objects by specifying a "physical" identifier:

L:P:D

or by just a "logical" identifier:

P:D

In the latter case the registry needs to be consulted to find out
where the object may be found (i.e., which service resource carries
the service).

It is important that copies of any project at different locations
honor the dataset identifiers issued by the original project, though
it should be allowed to develop extensions.

"Not available" will be a valid response and it should be possible for
the client to indicate that the request be forwarded to a more
complete copy of the service if not available at this location.


Note that it is the existence of mirror services and the ability to
move services from one resource to another that leads to having to
define three, rather than two sub-identifiers.

Note also that the identifier does not replace a URL.  However, it
allows querying the service (resource) for the URL that would give
access to the object identified by the identifier.
That may make it easier to handle obsolete identifiers.

Hope this makes things clearer,

  - Arnold


--------------------------------------------------------------------------
Arnold H. Rots                                Chandra X-ray Science Center
Smithsonian Astrophysical Observatory                tel:  +1 617 496 7701
60 Garden Street, MS 67                              fax:  +1 617 495 7356
Cambridge, MA 02138                             arots at head-cfa.harvard.edu
USA                                     http://hea-www.harvard.edu/~arots/
--------------------------------------------------------------------------