On Wed, Aug 15, 2012 at 03:48:13PM -0400, Aaron Rubinstein wrote:
> >> Aaron: I agree with Antoine. I like the distinction between SES (which equals
> >> rdf:Datatype) and something a looser, maybe for mapping from one serialization
> >> to another such as from MARC to ISBD. I think this should be separated from the
> >> notion of an SES.
>>
>> Should we consider mapping between serializations and mapping between
>> different data models and schemas to be separate functions? In library
>> jargon, the latter is called "cross-walking" and it often changes more than
>> the serialization -- there can be major modifications to underlying data
>> structures in order to accommodate different community models.
>
> This is an excellent point and I believe our thinking clarified a bit on this
> as the call developed. As I understood things, we discussed having a tightly
> defined SES that is an rdf:Datatype and something else, probably the DSP,
> that serves as a "serialization profile" that explicitly does not "crosswalk"
> but defines a data model that can be serialized into different formats. In
> other words, the DSP provides a guide or specification for serialization into
> specific formats, not from one data model to another.
A DSP can provide a _target_ for cross-walks. The target DSP itself may be
perfectly well-specified, and straightforwardly expressible in RDF triples,
even if the cross-walk to get there is of necessity messy, lossy, or incomplete.
A DSP can be a _source_ for cross-walks to the extent to which a straightforward
transform ('serialization') can be defined.
Tom
--
Tom Baker <[log in to unmask]>
|