On 2/29/12 6:23 AM, Jon Phipps wrote:
> Despite the 'baggage', I'd hate to see us throw out useful abstractions
> like 'Syntax Encoding Scheme' and even 'Vocabulary Encoding Scheme'
> _unless_ we can come up with equally abstract and precise but more
> colloquial terms.
I'm not suggesting that the abstractions or the terms be thrown out. I'm
suggesting that the data gathering phase should be action-based and
user-based, and that the abstractions should be derived from and shown
to support the user-based view. I see it was a work flow issue for the
group. I also see it was a way to make sure that the abstract model
serves actual user needs. This is hard for people to do because DCAM
already exists so the notions are already in people's heads. I suggest a
"willful forgetting" for a short while so that user needs can be
gathered. Then the abstract model can be postulated and the use cases
can serve to make sure that all user needs are met. Mixing the abstract
model with the use cases at this point pollutes the waters, so to speak,
because it is mixing proposed user tasks with pre-determined solutions.
> Forgive me if I'm stating the obvious, but I have observed when watching
> the use cases being developed for LOD-LAM, that they tended to boil down
> into a single metadata process flow use case:
> 1. Create data
> 2. Validate created data against an organizational/domain standard
> 3. Store valid data locally
> 4. Use locally (index, view, search, browse)
> 5. Disseminate/publish inter-organizationally (rss/atom, oai-pmh,
> linked data)
> 6. Aggregate inter-organizational data
> 7. Validate aggregated data against an organizational/domain standard
> 8. Store valid aggregated data locally
> 9. Use locally (index, view, search, browse)
> 10. ... Rinse/repeat
> Regardless of the specifics entailed by these processing and management
> steps, the use cases all end up repeating this pattern at some level.
> It's steps 2, 5 and 7 that the DCAM addresses:
> * What resources is our organization describing?
> * What finite list of properties do we use to describe them?
> * What constitutes valid data for each of those properties in the
> context of our organizational needs?
> * How can we describe those resources to consumers of our data about
> them when we publish it outside of our organization in a way that
> they will understand?
> The value of an abstract syntax (like SoDC or dc-text) for describing
> resource description validation and semantic constraints is its
> independence from implementation details. A metadata record is always a
> finite list of properties/statements; cardinality is cardinality
> regardless; a finite list of valid terms can be defined by a list of
> literals, or a location/identifier of that list, or a list of
> identifiers; a valid date must be a unix timestamp or it must be w3c; etc.
> Whether these constraints are expressed in RDFS/OWL or RIF or
> schematron, whether the data is stored or published in XML or RDF,
> shouldn't matter to the DCAM in exactly the same way that they shouldn't
> matter to the DCAP.
> Karen's generic list of things I 'want to do' is a perfect example of this.
> On Tue, Feb 28, 2012 at 8:32 PM, Karen Coyle <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
> On 2/28/12 3:09 PM, Thomas Baker wrote:
> Many of the patterns on  -- in my reading -- point to a
> constraint language.
> I wish I could be on this call, but I'm on another one. I do want to
> say that an example like:
> "A description with a single statement, which uses a literal value
> with a Syntax encoding scheme."
> is written in terms of an assumed solution rather than a user task
> which may lead to a solution. If you've already got things called
> "literal value" and "Syntax encoding scheme" you have started with a
> fair amount of baggage. I would suggest expressing the examples in
> terms of things that people want to do (actions) rather than
> particular solutions ("syntax encoding scheme") and see what
> solutions are needed.
> And example of 'want to do':
> - give a plain text description of something (the title of my book
> is "")
> - give a description of something using a controlled vocabulary of
> terms that are listed here ("red" "yellow" "blue")
> - give a description of something using a controlled vocabulary that
> exists as a list somewhere and can be identified (...)
> - express a value that is a known data type (e.g. currency)
> - describe something that has multiple parts, all of them plain text
> (RDA publisher statement, made up of place, publisher and date)
> - etc.
> If so, does a constraint language need to be based on an
> abstract model the way
> DSP is based on DCAM? What, then, are the requirements for DCAM?
> Rather than jump straight to these abstractions, I'd like to use
> examples to
> clarify the basic requirements.
> Karen Coyle
> [log in to unmask] <mailto:[log in to unmask]> http://kcoyle.net
> ph: 1-510-540-7596 <tel:1-510-540-7596>
> m: 1-510-435-8234 <tel:1-510-435-8234>
> skype: kcoylenet
[log in to unmask] http://kcoyle.net