(Apologies for crossposting to METS, PREMIS, MIX)
Sample METS document of a digitized monograph for comments.

The National Library of Finland is revising the specification for
digitized monographs to include PREMIS metadata and we are requesting
feedback a sample METS document.

The sample document and feedback received will be used to create a METS

We hope to receive feedback by 28th of August. Feedback can be given
via mail list or directly by email. As an incentive we will be giving
out small gifts to three randomly selected persons providing feedback.

	Jukka Kervinen				
	[log in to unmask]	
	System Analyst				
	National Digitisation Center	

	Karo Salminen
	[log in to unmask]
	System Analyst
	Library Network Services

	National Library of Finland


Brief summary of the sample document.

The METS document describes a digitized monograph with OCR and
structural markup performed with CCS docWORKS. It is intented to serve
as both the access and the preservation functions. Main emphasis in
PREMIS has been in recording events. Rights metadata is still under
discussion. The basic METS structure comes from schema defined
by MetaE-project and CCS.

The digital object consists of:
- master images in JPEG2000 lossless format
- access images in JPEG format, rescaled to lower resolution and
converted to sRGB
- thumbnail images in JPEG format
- OCRd text in ALTO XML format
- PDF with hidden text layer
- METS xml
Representations (PREMIS):
- repr1: master jpeg2000 images + alto files (+ Physical and Logical
- repr2: access jpeg images + alto files (+ Physical and Logical Structmap)
- repr3: thumb images
- repr4: PDF
METS file contains two dmdSecs with bibliographic record of the digital
in MARCXML and MODS formats. MARCXML record is derived from the analog
original's record which is also stored in <sourceMD> of the main amdSec

There also is a <dmdSec> for each chapter and illustration in MODS
format. Identifiers are NBN based URNs.

METS file contains one amdSec with id main-amd for administrative
metadata that applies to the whole object:
- rightsMD section
- sourceMD section with bibliographic record of the analog original in
MARCXML format.
- digiprovMD sections describing events (mainly document processing
steps) and agents that created the object and apply to the whole object
using premis:event and premis:agent schemas. The events in the sample
document are based on steps of the processing software used and will be
different when another software is used.
For each file in the METS package there will be a separate amdSec that
contains technical metadata in <techMD> and event metadata in
<digiprovMD>. <techMD> contains a premis:object with MIX and JHOVE
output in objectCharacteristicExtension.
Agents are referenced from main <amdSec>
Filegroup contains five <fileGrp>s for image, alto-xml and PDF files.
Document has Physical and Logical structMaps detailing the structure.
There are also additional structMaps for different representations.

Identifier types used in PREMIS are fictitious, used only for
demonstration purposes. Date fields do not contain actual values.


All comments regarding the sample are welcome. We identified a need of
feedback on the following points:


- MODSQ-01: MARC21slim2MODS3-3.xsl translates
<MARC:datafield tag="856" ind1="4" ind2="0">
<MARC:subfield code="u">
to <MODS:identifier type="uri"></MODS:identifier>.
Would <MODS:location>
</MODS:location> be more appropriate?
- Are there any significant gaps from long term preservation point of view?

- Do you consider saving JHOVE output in full reasonable?
What is the practice in existing PREMIS implementations?
(Sample includes JHOVE output for master image and PDF in

- Currently the content of representations are described in additional
structMaps. Do you find flaws in this implementation?

- What are your thoughts in general on recording events related to
digitization in PREMIS?

Specific (use find to locate from XML)

- PREMISQ-01: edtfSimpleType: We would like to record datetime range but
it is not accepted per v2.0 schema as shown example
(20050705T0715-0500/20050705T0720-0500) in the PREMIS data dictionary

- PREMISQ-02: We propose adding a new semantic unit to Agent entity,
agentNote to PREMIS data dictionary. It might be used for describing
software and its components in detail.

- PREMISQ-03: We record source relationship for the source image(s)
created temporarily for processing. Do you find this reasonable?
We also record the barcode of the scanned book in the scanning event

- PREMISQ-04: Should METS IDs and PREMIS identifierValues be identical?

- PREMISQ-05: Comment: According to PREMIS-v2-0.xsd
<premis:objectCharacteristicsExtension> can not be repeated in
This means that to have more than one external schemas for a file, in
this case a image file with MIX and JHOVE metadata,
<premis:objectCharacteristics> has to be repeated with identical

The Data Dictionary seems to indicate that repeating
objectCharacteristicsExtension should be possible.

- MIXQ-01: <mix:formatVersion> Is there a format version for JPEG 2000?

- MIXQ-02: Colorspace and ColorProfile with embedded scanner profiles
We will be using embedded device ICC profiles in the archive images.
What would be the appropriate way to indicate that in <mix:colorSpace>
and <mix:ColorProfile>.
     - colorSpace in master images with embedded profiles? RGB,
DeviceRGB, ICCBased?
    - Can embedded icc-profiles be marked in MIX?
- MIXQ-03: Targets. In our planned workflow targets (resolution,color)
are scanned once a shift and a separate digital object will be created
for the scans. These target objects are then referenced from MIX.
Recording multiple targets in MIX seems problematic. While the NISO spec
allows for listing of several targets validating against schema  gives error
Instead of trying to reference several targets contained in the target
object in MIX  would it be feasible to just reference the target object
- MIXQ-04: Archived master images are 2nd generation images ie. cropped
and deskewed versions of the scanner originals. Should the scanner
details be nested in PreviousImageMetadata? There are two versions of
MIX in the sample document. One with scanner data in the main data and
another with scanner data in PreviousImageMetadata.