I think I fall somewhere between Rob and Ed.
First, note that although I appreciate the nods to my work on MARC, I
wasn't involved in the creation of this vocabulary. I'm not yet
convinced that a level 0 of this nature is useful, but until we have
levels 1+ this remains unknown. I think that there may be a starting
point that might be useful for various transformation activities, but
I'm not at all sure that this is it. (Actually, Ian Davis had started
a level 0 or 1 that looked promising to me.) 
I could imagine a useful level 0 that would be a pass-through from
MARC to something else, and that in itself would never see the light
of day. I agree with Rob and Ed that providing a round-trippable
MARCRDF could be deceptive, in the same way that MARCXML has tricked
people into thinking that we now have a version of MARC that has the
extensibility qualities of XML. We should be careful that these
intermediate steps do not get mis-characterized as solutions.
As I understand it, this Level 0 approach is an atomizing of the
subfields of MARC, with each individual subfield/indicator combination
getting a URI. I'm having a hard time wrapping my mind around how this
is then transformed into levels 1 and 2, and am not sure from this
posting what features those levels have nor what the final goal is.
More discussion of the overall approach would be very useful.
I would request that the authors provide a description of how this
level 0 was developed and the decisions that were made, since it is
not easy to understand from viewing the mass of elements in the
registry, and how it is anticipated that this feeds into their vision
of an end goal. My qualms are that the inherent *structure* of the
MARC record, with tags that gather subfields and provide meaning, has
to be addressed, and that is a level 1 or 2, so we don't see any
evidence of that here. In the interests of science and "running code"
this multi-level approach now needs to be tested, or at least put in
the context of a model of the next level or two.
 http://blog.iandavis.com/2005/12/22/marc-transliteration/ although
the blog post is broken. Somehow i managed to see it all some time in
Quoting Ed Summers <[log in to unmask]>:
> I agree in principle with Rob. However, I think Level0 could be a
> useful for a couple reasons:
> * Level0 could be a common vocabulary for folks who want to pour
> extant MARC data into an RDF triplestore, so that they can then use
> SPARQL and inferencing to massage the data into another more
> expressive vocabulary, like RDA, FRBR, etc. People who primarily use
> relational, xml or nosql databases probably won't care about this too
> much...but that doesn't mean RDF triplestore users won't does it?
> * Converting MARC to Level0 RDF could establish URIs for the
> bibliographic items that are being described. RDF triples are
> statements about a subject resource, which is identified with a URI.
> So Level0 would necessarily have to hang RDF assertions off of a URI.
> This might seem like a small thing, but it actually would be quite a
> significant improvement in the current state of affairs, where our
> bibliographic resources lack an identity on the Web. This assumes that
> HTTP URIs were used of course :-)
> As Rob correctly points out, the big danger here is that people
> convert their MARC to Level0 so that they can check the Linked Data
> box on some RFP; and that all the work that has gone into higher level
> vocabularies (by some of the same people as who created Level0 I might
> add) is ignored.
> On Tue, Sep 13, 2011 at 4:35 AM, Rob Styles <[log in to unmask]> wrote:
>> Hi Gordon and others at MMA.
>> First of all can I say how great it is to have still more folk looking at
>> the potential of RDF and Linked Data for libraries.
>> Levels 1 and up sound like they might be a really interesting contribution
>> to the debate.
>> Many people at Talis, Library of Congress, Dublin Core and throughout the
>> Linked Data community have worked hard to discuss approaches to representing
>> bibliographic data that will benefit libraries and their participation in
>> the wider web. Much of this effort has been contributed altruistically and
>> great strides have been made with several national libraries and some
>> Several years ago the library community had the opportunity to publish data
>> in XML for the XML community. The result was MARC XML. Unfortunately,
>> because the translation from MARC was one-to-one many in the library
>> community bought into it and other more useful XML representations have all
>> but died as a result.
>> Your Level 0 risks the same. At a time where the library community needs to
>> understand the meaning in their data and let go of a language (2XX etc) that
>> came from the computing constraints of the late '60s the vocabulary you have
>> published is not helpful. It provides a simple cop-out to understanding, and
>> doing, something genuinely useful.
>> I grant you, it has a use-case. Round-tripping between MARC and Linked Data
>> will be something that many people will ask for, yet very few will actually
>> need. We risk that people who don't know better will convert MARC to this
>> vocabulary and think they have data that the Linked Data community will find
>> This Level 0 vocabulary has the potential to lead people down an unhelpful
>> cul-de-sac and take time, effort and understanding away from genuinely
>> useful work such as that from LoC, BL and DC.
>> I beg you — for the sake of libraries everywhere — delete it and never speak
>> of it again!
>> Rob Styles
>> Senior Technical Consultant
>> On 13 September 2011 09:10, [log in to unmask]
>> <[log in to unmask]> wrote:
>>> Elements of MARC21 bibliographic standard represented in Resource
>>> Description Framework by Metadata Management Associates
>>> Metadata Management Associates (MMA) is pleased to announce a new resource
>>> for the library community: The MMA version of MARC 21 elements and
>>> vocabularies in RDF, hosted by the Open Metadata Registry (OMR)
>>> [http://marc21rdf.info/] . Given the need for libraries to move beyond the
>>> MARC standard, and the desire for increased innovation as we move toward a
>>> successor, we felt this was a good time to make this data available to all.
>>> The MMA version of MARC 21 in RDF has been exclusively developed
>>> by Metadata
>>> Management Associates, and has not, as yet, been shared with the Library of
>>> Congress, although the announcements on MARC [
>>> http://www.loc.gov/marc/transition/ ] made it even more desirable
>>> to move in
>>> this direction.
>>> So why did we spend the time to do this, given the intention to leave MARC
>>> behind? Most importantly, we want to make it easier for innovators to
>>> ‘play’ with this data, and to have URIs to use when they do so. We
>>> ourselves want to include MARC21 elements in our research into semantic
>>> mapping between bibliographic namespaces, and to inform proposed work on
>>> representing UNIMARC in RDF. We also want to provide inspiration for those
>>> eager to experiment with the issues around the transition from MARC 21 to
>>> the new environment of RDA and other bibliographic standards, and
>>> trust that
>>> the reassurance that it WILL happen--and is not rocket science--will help
>>> everyone interested in participating in that future.
>>> What you see in the OMR now is what we’re calling ‘Level0’, the most basic
>>> loss-less way to transition MARC 21 data into the Resource Description
>>> Framework (RDF). It does not reflect the layers we think ought to be added
>>> on top:
>>> Level1 may contain properties that gather sub-properties with similar
>>> semantics from Level0.
>>> Level2 may contain properties and classes that represent aggregated
>>> statements composed of sub-properties from Level0 and Level1.
>>> Level3+ can contain properties and classes that represent broader-level
>>> semantics and provide equivalence mappings to other namespaces for
>>> bibliographic metadata, such as Dublin Core terms, ISBD, and RDA.
>>> Special thanks to Karen Coyle, whose work on the analysis of MARC 21 (most
>>> recently seen in a Code4Lib Journal [
>>> http://journal.code4lib.org/articles/5468 ]) inspired us greatly and whose
>>> questions pushed us in some important directions. Karen’s insights
>>> and ideas
>>> will be far more visible in Level1 and subsequent levels, when we build on
>>> the basics.
>>> This initiative is also intended to inform the proposal to develop an RDF
>>> representation of MARC 21’s cousin, UNIMARC, presented at IFLA this year [
>>> Please note that not all MARC21 elements are currently represented in the
>>> OMR. Some of the lesser-used tags in 00X-8XX have not yet been registered,
>>> and we are still looking into 76X-78X Linking entry fields.
>>> We welcome feedback on this effort. Information on errors and specific
>>> issues is best communicated via the “FEEDBACK” links on all OMR pages.
>>> We’re happy to participate in discussions on the DCMI/RDA Task Group
>>> discussion list [ [log in to unmask] ], or anywhere else--just
>>> be sure we
>>> know that you’ve posted something, and we’ll respond.
>>> Those of you headed for the DC-2011 conference are likely to see us
>>> presenting information on our new mapping initiative, and demonstrating how
>>> the MARC21 data in RDF supports useful mapping from MARC21 to RDA. We’ll
>>> post slides and links after that conference.
>>> Metadata Management Associates
>>> Diane Hillmann
>>> Jon Phipps
>>> Gordon Dunsire
>> Rob Styles
>> +44 (0) 7971 475 257
[log in to unmask] http://kcoyle.net