For those not able to attend the METS meeting prior to
the DLF Forum in Pittsburgh, I'm posting the following notes
from a very interesting and valuable meeting.  Several items 
of business have come up that will need to be addressed in the 
short and medium term.

The most immediate of these is finalizing version 1.0 of
the METS schema.  All of the changes reflected
in the gamma version of the METS schema currently available
through the Library of Congress web site were approved
and accepted by the group at the meeting.  Several other changes 
have been proposed, which I will summarize below.  You may consider
this message an official call for comment on these proposed
changes.  If we see no significant disagreement on these
changes in the next two weeks, they will be implemented 
as part of the final version 1.0, which will then go to
DLF for formal review for possible endorsement
as a DLF recommendation.  The proposed changes are as follows:

1. Alter the definition of structMap so that only one root <div> 
element is possible for each structMap. [NB: a single root <div> was 
the original intent for this element, and conforms to the DTD for MOA2; 
the ability to have multiple root <div> elements was a mistake
introduced by myself in converting from DTD to XML Schema.]

2. Add the following roles to the controlled vocabulary for the ROLE 
attribute on the agent element in the METS header: CUSTODIAN, 
RIGHTS OWNER, OTHER.  Additionally, add an OTHERROLE 
attribute to specify a particular role outside the controlled 
vocabulary employed in the `ROLE' attribute when a ROLE of 'OTHER' is 

3. Add a new, major subsection to METS directly beneath the root 
METS element for referencing software used with the entirety or 
portions of a METS document.  This element was tentatively dubbed 
`behaviorSec' at the meeting, and will operate similarly to the draft 
that Sandy Payette proposed for enabling support of FEDORA objects 
using METS.  Individual 'behaviors' will be specified using a format
specified using extension schema, and any behavior may be linked to 
one or more sections of a METS document.  There are no restrictions 
on which portion of a METS document a behavior may be linked with.

4. All attributes of type "date" will become type "dateTime".

5. Rick Beaubien from UC Berkeley pointed out that the current content
model for mdWrap (which allows arbitrary XML by specifying the `##any' 
namespace for subelements) only allows a single root element beneath 
the  mdWrap element (although arbitrary amounts of subelements beneath 
that root element are allowed), and that it would be preferable from 
Berkeley's perspective to allow more than one subelement immediately 
beneath an mdWrap from a non-METS namespace.  Jerry McDonough and Rick 
Beaubien will investigate and report back to the list whether this is 
possible without major revisions to the schema, and if so, this will be 

6.  The documentation for descriptive metadata and administrative 
metadata sections will be clarified to indicate that metadata within 
these sections can be referenced from either the structural map or the 
file section of a METS document.

7. The CHECKSUM attribute, currently located on the FContent element, 
will be relocated to the file element so that it can be used with files 
internal or external to the METS document (i.e., FLocat or FContent).

8. The DMD attribute on <div> will be renamed to DMDID for 
consistency's sake.

9. There was a great deal of discussion about how best to store 
technical metadata that pertains to data files, and when it was best to 
store it in an administrative metadata section (typically, when the 
same data recurs for many files) and when it was best to store it on a 
file element (typically, when there's a great deal of variation in the 
information from file to file).  UC Berkeley expressed a hope that the 
dimension attributes used on file elements in the MOA2 DTD might be 
restored in the METS schema, as such information tends to be specific 
to individual files, and they'd rather *not* store such information in 
an administrative metadata section separately from the file.  Jerry 
McDonough and Rick Beaubien will investigate and report back to the 
list whether it's possible to employ namespaced attributes 
that do NOT belong to the METS name space on a METS element.  If so, 
then organizations which want to store information about files which 
varies significantly from file to file might be able to do so without 
requiring a formal change to the METS schema.  Assuming this is 
possible, documentation regarding how to employ namespaced attributes 
for this purpose will be added to the METS schema.

In addition to the schema changes, a variety of administrative
issues were addressed at the meeting.  Probably the most important
of these, as mentioned already, is that when the next version
of the METS schema is finalized, it will be sent to the DLF
for formal review and possible endorsement.  At the meeting,
we decided we would like to try to conclude a formal review
process within three months of submitting the schema to DLF.
I discussed this with Dan Greenstein at the Forum, and he
believed that schedule was reasonable.

A great deal of discussion was also devoted to the organizational
structures needed to continue developing METS when version 1.0 is
completed.  The group decided that it would be best to create
an editorial board responsible for further technical development
of METS.  I will work with RLG to help form such a board and 
serve as editor-in-chief for the time being.  All meetings of this board
will be open to the public, and the board will actively solicit input 
from the METS mailing list and other sources about possible changes to
the METS schema and other technical developments.

It was decided that anyone who wishes to should be able to
register an extension schema for use with METS.  Registered
schema will be made available through the METS web site at
the Library of Congress.  The METS editorial board may decide
to endorse particular extension schema for use with METS
when multiple schema for the same metadata have been registered
(e.g., multiple schema for Dublin Core).

There was also discussion of what exactly needs to be registered
in terms of controlled vocabularies for use in METS, particularly
with regards to the <div> and <mets> TYPE attributes, and the
<mets> PROFILE attribute.  While the group did not see a pressing
need for controlled vocabulary for the TYPE attributes, controlled
vocabulary for the PROFILE attribute was seen as necessary for
supporting the creation of software to work with METS objects.
The editorial board for METS will need to immediately turn its
attention to the issue of formalizing Ďprofilesí for METS objects.

Possible formal standardization of METS was discussed, either
through NISO or the World Wide Web Consortium.  While the
group could definitely see benefits to this, it was felt to be somewhat
premature.  After version 1 of the METS schema has been reviewed
by DLF and there has been some opportunity for the METS editorial
board to receive feedback from groups implementing systems using
METS, the editorial board will revisit the question of pursuing
formal standardization of METS.

To the extent possible, the group decided to try to divvy up some
of the work of developing shareable extension schema for administrative
metadata, based on work in progress.  Michigan State Univ. has
already done work building upon the extension schema for technical
metadata released by the Library of Congress; it was decided that MSU, 
LC and Harvard should try to coordinate their work on developing
extension schema for technical metadata for audio and video resources
and make their work available at the METS web site.  MSU, Harvard and
NYU will work on still image technical metadata extension schema.
NYU will work with Harvard on developing a technical metadata extension
schema for use with text materials.  NYU will work with Cornell and
UVA on extension schema for intellectual property rights, with 
Cornell/UVA focused on rights as they relate to software behaviors 
(essentially encoding policy with regards to METS objects use/display) 
and NYU focused on broader intellectual property rights issues (i.e., 
integrating  XrML and ODRL work with METS).  We would like to see 
enhanced versions of extension schema for all of the administrative 
metadata sections available through the METS web site by the end of 

There was a great deal of discussion about software development for
use with METS; there has been an impressive amount of work already
done on this.  Everyone who has been developing software tools or
extension schema for use with METS are encouraged to contact
Morgan Cundiff at LC so that pointers to information about these
efforts can be listed at the METS web site.

I think itís appropriate at this point of METSí development
to officially recognize a few people for their contributions.  In 
particular, I would like to thank Rick Beaubien, Morgan Cundiff, Carl 
Fleischhauer, Mackenzie Smith, Sandy Payette and Merrilee Proffitt for 
all of their  work in helping to turn METS into something useful, and 
for helping make my life a lot easier.  Dan Greenstein deserves a big 
nod for getting this particular ball rolling in the first place.  And 
Iíd also like to thank all of you out there on the list for 
participating in the METS initiative and helping push the digital 
library world towards a solid and shareable model for our metadata.

A revised METS schema should be available at the METS web site
in about three weeks, assuming the changes listed above are
non-controversial.  Once this version is finalized, I will talk to
Dan Greenstein about initiating a formal DLF review and start
working with RLG to compose an editorial board for further
work on METS.  For the moment, if you have any comments
on the proposed changes, please send them along to the METS