On Wed, 19 Oct 2005, Brian Tingle wrote:

> On Wed, 2005-10-19 at 16:35, jeroen bekaert wrote:
> > Changing mandatory/optional occurence constraints based on METS related
> > issues is clearly an 'interoperability with other schemas' problem. Its
> > sole purpose is the trouble-free use of PREMIS in conjunction with METS
> > (only).
> I'm a little lost as to why this is a METS only issue.  The redundancy
> issue does not crop up in any other context?
> If only a few elements are mandatory, then maybe its moot.
> > 
> > It should be clear that mandatory/optional contraints on PREMIS
> > elements/attributes should be dictated by the abstract PREMIS data
> > dictionary only.
> That does make sense to me, but it was suggested at the top of this
> thread that the definition of "mandatory" in the context of the data
> dictionary is not the same as the definition of "mandatory" in XML
> Schema. 

That is true that the definition of mandatory in terms of the PREMIS data
dictionary is different from mandatory in terms of an XML schema (from
chapter 2 introduction to the data dictionary):

"Obligation: Whether a value for the semantic unit is mandatory (if
applicable) or optional

A mandatory semantic unit is something that the preservation repository
needs to know, independent of how or whether the repository records
it. The repository might not explicitly record a value for the semantic
unit if it is known by some other means (e.g., by the repository's
business rules).  "Mandatory" actually means "mandatory if
applicable."  For example, an identifier for a bitstream is mandatory only
if the repository manages data at the bitstream level.  When exchanging
PREMIS-conformant metadata with another repository, values for mandatory
semantic units must always be provided.

Values for optional semantic units are encouraged but not required.

If a container unit is optional, but a semantic component within that
container is mandatory, the semantic component must be supplied if and
only if the container unit exists.  That is, if a value for any of the
optional or mandatory semantic units in the container is supplied, a value
for all of the mandatory semantic units in the container must be

The working group spent a lot of time on mandatory vs. optional and came
up with this definition because of the variety of practices in and
implementations of preservation repositories. For instance, a repository
may use an object identifier that is unique in its own system, so although
objectIdentifierType is a mandatory element, it does not have to be
explicitly recorded within the repository. However, if the metadata and/or
object were exchanged, the type would need to be generated on

> >  After all, isn't this why people define an abstract
> > model? See also the email sent by Ryan Chute on July 11 on this very list.
> Is there an archive of this list?  I was not on it in July 11. 

Yes, see:

You will have to set up a password to access; it will prompt you to do so.


^^  Rebecca S. Guenther                                   ^^
^^  Senior Networking and Standards Specialist            ^^
^^  Network Development and MARC Standards Office         ^^
^^  1st and Independence Ave. SE                          ^^
^^  Library of Congress                                   ^^
^^  Washington, DC 20540-4402                             ^^
^^  (202) 707-5092 (voice)    (202) 707-0115 (FAX)        ^^
^^  [log in to unmask]                                          ^^
^^                                                        ^^