Carl, interesting you should ask.  When NAMTC first began looking at the issue of standardizing metadata, they really only identified bibliographic data as metadata.  Part of my proposal back to NAMTC was to point out that for digital objects such as this metadata has grown by necessity past bibliographic data (something I believe everyone already working in this field had already identified and in fact done pretty extensive work on).

I identified, after some initial research into METS, MPEG 21, and other standards, 6 types of metadata that this particular group (the educational video market) might be interested in:

-- Bibliographic

-- What I termed concordance, what I believe you are terming descriptive metadata - that is, data that describes elements of the content and is aligned to the content in a spatio-temporal way.  That is, a description of who and what is on screen at 5:43 into the video, etc.  We in fact are using Virage (interesting you should identify that) as our indexing package, and of course are supplementing its automatically defined data with human entered data.  Which, I am sure you are aware, is an excessively long and difficult process.

--Structural, as you say, to keep actual digital files linked together into a single digital work.  Recently, after reading Priscilla Caplan's "Metadata Fundamentals for All Librarians" I have begun to see the need for super-structure data as well - linking different digital works together into collections or series based on use, topic, or other reasons the original content creator linked various works (I believe here is also where FRBR has an impact).

-- Technical, as you identify below.  NAMTC is also trying to standardize formats for digital video for content creators to make available (for instance, 384 Kbps MPEG 4 as one standard, 1.15 Mbps MPEG 1 as another), to try and cut down on the disparity of digital video formats out there.  This type of work is really already being tackled however by groups like the ISMA (http://www.isma.tv/) and MPEG themselves.  I am sure we are all aware however of the trouble MPEG 4 has had getting off the ground.

-- Administrative, including rights management.  As a content producer, this is clearly very important to us.  Though I recently read a analysis of how this fits into the metadata world as part of the MPEG 21 statement of intent that I thought summed up the problem well - consumers of content are only going to accept DRM and IPRM if it does not inhibit their fair use of the material.

-- Pedagogical, which would be where a standard of metadata would start to cross over with standards of teaching packages such as NISO's Learning Object Model standard (formally IMS's standard).

I did not identify preservation data initially because as current content producers we are not dealing with any content in a format in threat of deterioration in the near future.  However, I know libraries and archives are using digital as a preservation technique and any standard we would use as a current content producer we would want to fit into the needs of the library, including working with archive management systems that expect some level of preservation data.  We would like to record the process of how the files were created as well, as we feel it is important data for a customer to have to understand the quality of the source material the quality assurance used to create the digital material.  Similar to CD's in the consumer market, it is easy for the consumer to think that because the format is digital quality it makes all the sound the same quality, when in fact the process used to create the CD is the most important factor in quality.

In my initial proposal back to NAMTC, I identified bibliographic, concordance, and structural as data that should be immediately available (due to this industry's long use of MARC, for instance) or of the most immediate value.  The other types are going to be a little harder to achieve and create I believe, which is to not to say that in the end all of the types are not of important value.  In fact, I identified these types after seeing the work already done on METS and MPEG 21 because it is clear from the work all of you have done that all of the types are required to really make management and use of a digital collection possible.

Hope that answers your question.
_________________________________________

Thank you for the intriguing posting.  Some of us are also looking forward
to managing video in digital-file form and share your interest in the
general topic of appropriate metadata.  But help us with a clarification:

-- To what degree is your interest in "bibliographic" or
"descriptive" metadata, the kind of information that supports resource
discovery?  If in this category, does your interest include the output of
logging software, e.g., Virage?

-- To what degree is your interest in what digital library folks call
"structural" metadata, the kind that is needed to, say, keep a
multi-file/multi-part object wrapped as a package, for delivery or for
archiving?

-- To what degree is your interest in the kinds of technical metadata that
might describe the characteristics of files in an object package, e.g.,
"this is MPEG-4 Simple Profile, with a picture size of 360 lines and 240
samples per line."

-- To what degree is your interest in metadata needed to support systems
that manage access to files, e.g., rights, restrictions, or payment data?

-- To what degree is your interest in the kinds of "preservation metadata"
that is presently being cooked up by RLG/OCLC
(http://www.oclc.org/research/projects/pmwg/)?

Best wishes,

Carl Fleischhauer
Project Coordinator
Office of Strategic Initiatives
Library of Congress
Washington, DC 20540-1330
202-707-3979