Re: Keeping R2RML free of Direct Mapping dependency (ISSUE-25)

On Tue, 2011-04-26 at 09:08 -0500, David McNeil wrote:
> I see at least two usage scenarios:
> 
> 1) Just take an RDB and make its contents accessible as RDF. In my
> opinion the Direct Mapping satisfies this need. It is not necessary to
> tie R2RML to the Direct Mapping to achieve this. (This does not
> address the case of mapping a _changing_ RDB schema to RDF, but from
> my perspective that seems like a distant edge case that we do not need
> to address directly in the 1.0 version of the specs.)

s/the Direct Mapping satisfies this need/any tool based on the Direct
Mapping (definition) satisfies this need/

Otherwise, I entirely agree. This is why I wanted to call the Direct
Mapping _Direct_ Mapping and not _Default_ Mapping.

> 
> 2) Craft an R2RML mapping to expose a view of an RDB as RDF. This
> could be targeting a pre-defined domain ontology or crafting a domain
> ontology as part of the mapping exercise. In either case, I think the
> R2RML mapper needs explicit control over which RDB entities are
> exposed.
> 
> If the user wants a hybrid of these two models then they can generate
> the Direct Mapping for an RDB and then replace parts of it with a
> hand-crafted R2RML mapping.

The user never needs to define any Direct Mapping, but the Direct
Mapping mappings gives him a Direct Graph, materialized or not.

So another approach that I want to mention is:
1. the user generates the Direct Graph
2. the user targets "a pre-defined domain ontology" based on the Direct
Graph using RDF2RDF technologies like RIF, SPARQL CONSTRUCT, etc.

1. and 2. can of course be used as a declarative approach (with no
materialization) in conjunction with reverse-mapping techniques:
* the Direct Mapping becomes SPARQL2SQL
* RIF / SPARQL CONSTRUCT becomes SPARQL2SPARQL

Alexandre.

> 
> -David

Received on Tuesday, 26 April 2011 14:44:06 UTC