Difference between revisions of "Talk:ESciDoc Content Model Object"
Jump to navigation
Jump to search
m (→Brainstorming) |
|||
| Line 39: | Line 39: | ||
**might be more complex for self-archiving | **might be more complex for self-archiving | ||
**discussion | **discussion | ||
**inheritance might be good | |||
**revisit again, not initial implementation | |||
**templates can be used to define similar CModels | |||
*CModels define the behavior of Services (search, Relations, technical metadata-extraction) | *CModels define the behavior of Services (search, Relations, technical metadata-extraction) | ||
Revision as of 16:06, 24 November 2008
Agenda discussion 24.11.2008[edit]
Brainstorming[edit]
- Validation
- Classification of objects (typization)
- agreed by all
- CModel should be defined in XML (XSD, Schematron, RDF, RelaxNG, proprietary)
- agreed by all
- (Formal) structure of objects "festlegen"
- agreed by all
- CModels to their digital objects are what classes are to instances
- discussion, MRa point: to stick as much as possible to Fedora CMA and the new developments coming out from Fedora
- questioning why eSciDoc is "hiding" Fedora (besides more complex security model, support for aggregations and object graphs)
- example: OAI-PMH had to be re-worked (eSciDoc have more requirements on set definition and specific version of objects, not last versions, but released versions)
- CModels describe the semantic of instances (textual, context information, content category, metadata profile etc.)
- semantic defined as meaning of objects of certain CModel
- semantic is described with a certain formal structure
- semantic is described by syntax such as XSD, Schematron, RDF etc.
- Define type-specific methods
- all agreed there can be type-specific methods even though not always clear where/how
- discussion on Resource (Object manager) handler
- at end agreed that type-specific methods are not necessarily equal to Fedora disseminators (can be defined by users of the CModel i.e. for purpose of certain solution)
- Container handler is convenient even though it is apparently clear Items are very similar to containers
- independent from the implementation
- CModels should not contain implementation specific details
- why?
- E.g. CLARIN
- can contain link to e.g. WSDL, REST interface (as contracts for type-specific operations)
- Shall we go for WSDL, other description?
- resource behaves as a service (with contracts for it's own specific operations)
- evtl. other alternatives, but always "web service"
- interoperable
- related to Implementation non-specific
- MRA: CMX
- Namespaces for CM+Formats+MD-Profiles+URIs
- inheritance from CModels
- each cModel should be self-contained
- might be more complex for self-archiving
- discussion
- inheritance might be good
- revisit again, not initial implementation
- templates can be used to define similar CModels
- CModels define the behavior of Services (search, Relations, technical metadata-extraction)
- metadata profiles
- allowed relations
- life-cycle (status-transitions, initial status)
- CModel is versioned
- Templates
- File formats