For those of you attending the DLF Fall Forum next week, there will be a
discussion session on Tues. Nov. 6 from 5:00-6:00pm on Establishing Best
Practices for use of PREMIS within METS.

Below is some background and issues to be addressed during this session.
This is also available at:

The first draft of the best practices is at:

^^  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]                                          ^^
^^                                                        ^^


Questions for DLF Discussion session on: Establishing Best Practices for
use of PREMIS within METS
Nov. 6, 5:00-6:00 (Walnut Room, 3rd floor)

A draft best practices document was distributed to the METS and PIG lists
in August 2007. There was some discussion and comments about various
points. Moving this work forward is important for institutions
implementing PREMIS in a METS context and for the possible exchange of
METS documents between repositories.

Questions addressed in the draft are below with a brief summary of

1. Use of PREMIS schemas in METS sections
Most agreed with using premis:object in techMD, premis:event in
digiProv; premis:rights in rightsMD; premis:agent with appropriate section
(digiProv or rights).
Some may want to use mdRef instead of mdWrap, referring to metadata from
the appropriate section (using xPointer and xLink).
2. Number of sections to use	
Better to specify than leave it open for specific
implementations; repeating subelements under amdSec with use of ID/IDrefs
would be appropriate.

3. Use of PREMIS container
Most wanted to use the PREMIS container only if keeping all PREMIS
metadata together. This may apply to either mdWrap or mdRef.

4. Use of PREMIS with format specific technical metadata schemas
Ideally those elements defined in both PREMIS and format specific schemas
are only in one place  because of inconsistencies that may arise. There
was a suggestion to use xPath expressions to minimize
redundancies. Another suggestion was to have a premis element with IDref
to an mdSecType containing format specific metadata.
Note that the PREMIS Editorial Committee is considering a mechanism for
extensibility to keep format specific metadata together with PREMIS
metadata. When applied to METS that would mean optionally adding format
specific (e.g. MIX) metadata in the same section as PREMIS (or if using
mdRef in the same XML document).
5. METS structMap and PREMIS structural relationship elements
Most respondents suggested using the METS structMap with no redundant
PREMIS elements.

6. Other METS redundancies
Use PREMIS for attributes that are also defined in METS (e.g. CHECKSUM,
SIZE) with no redundancies. There was also a suggestion to have a pointer
(a new element or attribute, e.g. .valuePointer.) that could point to
another section of the METS document where the attribute is, usually using

7.  METS ID/IDREF and PREMIS identifier elements
Some thought using ID/IDRef is already well established in METS and should
be used. There was a suggestion to use the XML construct KEY and KEYREF
(although this approach did not seem well understood).

Issues to be addressed at the discussion session

1. Is it possible to even propose a best practices document without a lot
more experimentation? Are there advantages to waiting for some
experience? What would people do in the meanwhile?

2. Go over the issues and consider whether there is clarity on which
choices might be included in a best practices document (assuming the above
results in a desire for such a document).

3. Consider changes that might be needed to METS for PREMIS use.
Note that a request is going to the METS Editorial Board to include more
specific values for MDTYPE (to reference each of the separate PREMIS

4. Is it desirable to have the option of using PREMIS for all technical
metadata? i.e. an extensibility mechanism to include format specific
metadata within a PREMIS expression.

5. Using METS and PREMIS for compound objects when there is more than one
representation (in PREMIS terms) in one METS document for the intellectual
entity. Is the structMap or more? This was brought up on the METS
and PIG lists.