We at NLA have been looking more closely at relationships.

I noticed the problem of relatedEventIdentifcation being mandatory in
relationship, but found the email in the PIG list archive that says the
obligation needs to be changed (Rebecca's message to the PIG list dated
1 July 2005).  I have a couple of other comments to make about
relationships in the data dictionary.


Under relationshipType on page 2-62 it says "structural=a relationship
between parts of an object". That much is clear and accords with what
PREMIS says on page 1-8 and with what we understand i.e. structural
relationships are about how to put back together a digital object which
consists of more than one part or file. However the paragraph under
Derivation relationships on page 1-9 says "A structural relationship
among objects can be established by an act of derivation before the
objects were ingested by the repository ... " and "..They do not have
derivation relationships with each other, but do have a structural
relationship as siblings (children of a common parent)". It's confusing
to describe this as a structural relationship because the 'siblings' are
not part of the same digital object - they belong to different


On page 2-63 it says "is child of = the object is directly subordinate
in a hierarchy to the related object ..." and "is parent of = the object
is directly superior in a hierarchy to the related object ...", but it
doesn't say what the hierarchy relates to. In the paragraph (on page
1-9) referred to above, "parent" refers to the object from which the
"children" are derived, whereas on page 6-5 "children" is used to
describe components of a web site. In the former case the parent has a
"source of" relationship with the children; in the latter case the
children have an "is part of" relationship with the (parent) website. In
NLA's inhouse system (called Digital Collections Manager)  the term
"child" is used to denote "part of" at the Intellectual Entity level.
Because "parent" and "child" can be used in various contexts, I'd like
to suggest that the values "is parent of", "is child of", "has child"
and "has parent" be avoided in relationshipSubType and that more precise
terms such as "source of", "derived from", "is part of", "has part" be
used instead. 

We will be looking more closely at controlled vocabularies for
relationshipSubType and other PREMIS elements with this data constraint
in the near future.

Bronwyn Lee
Australian Partnership for Sustainable Repositories
National Library of Australia