No. It doesn’t work to argue about hypothetical analogies. Linked Data means all databases are connected (“linked”). As soon as you allow multiple formats, you might as well stick with web 2.0′s APIs. Whether RDF or not, there needs to be a single format (or at least a few, standard formats) that a technology such as SPARQL can parse through. As soon as computers can’t *automatically* understand the semantics of a piece of data, it is no longer the semantic web. As soon as computers can’t *automatically* link to and from a piece of data, it is no longer linked data.

If multiple formats was argued for by the community strongly enough, then I suppose we could use the idea of ontologies to have a way that SPARQL-like software could look up the way to interpret the software to find out how to interpret the implied triple. But wouldn’t that just be more of a headache than requiring everyone to use RDF? Front-end web designers (the thousands who will need to be convinced for this to really take off) are used to a one language per function paradigm (HTML for structure, CSS for style, Javascript for behavior). As soon as we give them choice, we’re just asking for frustration from all sides.

No. It doesn’t work to argue about hypothetical analogies. Linked Data means all databases are connected (“linked”). As soon as you allow multiple formats, you might as well stick with web 2.0′s APIs. Whether RDF or not, there needs to be a single format (or at least a few, standard formats) that a technology such as SPARQL can parse through. As soon as computers can’t *automatically* understand the semantics of a piece of data, it is no longer the semantic web. As soon as computers can’t *automatically* link to and from a piece of data, it is no longer linked data.

If multiple formats was argued for by the community strongly enough, then I suppose we could use the idea of ontologies to have a way that SPARQL-like software could look up the way to interpret the software to find out how to interpret the implied triple. But wouldn’t that just be more of a headache than requiring everyone to use RDF? Front-end web designers (the thousands who will need to be convinced for this to really take off) are used to a one language per function paradigm (HTML for structure, CSS for style, Javascript for behavior). As soon as we give them choice, we’re just asking for frustration from all sides.

The answer to your question lies in simply understanding that RDF is not the Atom re. graph data models and metadata.

The Atom is the Entity-Attribute-Value + Classes & Relationships (EAV/CR) data model [1].

The novelty of Linked Data meme lies in the power of the HTTP URI when applied to the EAV/CR data model.

Until that day arrives when we accept that computing, data structures, identity, data access, didn’t start with the Web, the entire concept of Linked Data will sadly remain mercurial to most

For what its worth here are some links:

1. http://en.wikipedia.org/wiki/Entity-attribute-value_model – EAV/CR data model
2. http://www.youtube.com/watch?v=WEj9vqVvHPc – really watch and listen .

Kingsley

The answer to your question lies in simply understanding that RDF is not the Atom re. graph data models and metadata.

The Atom is the Entity-Attribute-Value + Classes & Relationships (EAV/CR) data model [1].

The novelty of Linked Data meme lies in the power of the HTTP URI when applied to the EAV/CR data model.

Until that day arrives when we accept that computing, data structures, identity, data access, didn’t start with the Web, the entire concept of Linked Data will sadly remain mercurial to most

For what its worth here are some links:

1. http://en.wikipedia.org/wiki/Entity-attribute-value_model – EAV/CR data model
2. http://www.youtube.com/watch?v=WEj9vqVvHPc – really watch and listen .

Kingsley

I find this statement confusing: “Yes… but if, for the sake of argument, I happen to have ‘the idea of a triple’ in some other form, I may not want or need to convert.”

All RDF is, is a way of talking about linking URIs with URIs – in a form of triple statements. Everything else – N3, RDF/XML, turtle, etc. are just formats for writing it down.

When you understand that, Toby’s comment is simply correct, and nothing else needs to be said. If you have triples and URIs it is hard to come up with something different from RDF.

If someone comes up with something different, it will have to be so similar to RDF that arguments for using this (to be invented) approach versus RDF are not likely to be very substantial. And in any case, it will be quite easy to convert between it and RDF.

Irene

I find this statement confusing: “Yes… but if, for the sake of argument, I happen to have ‘the idea of a triple’ in some other form, I may not want or need to convert.”

All RDF is, is a way of talking about linking URIs with URIs – in a form of triple statements. Everything else – N3, RDF/XML, turtle, etc. are just formats for writing it down.

When you understand that, Toby’s comment is simply correct, and nothing else needs to be said. If you have triples and URIs it is hard to come up with something different from RDF.

If someone comes up with something different, it will have to be so similar to RDF that arguments for using this (to be invented) approach versus RDF are not likely to be very substantial. And in any case, it will be quite easy to convert between it and RDF.

Irene

Clearly your original thoughts where about RDF/XML the format. I discern that with clarity from your comments.

As I indicated in response to Ross, I can bear with current protraction of RDF model and data representation conflation problem; especially, since the version 1.0 of the RDF story is fully to blame for this confusion.

Today though, we do have a clear distinction between the model and its data representational formats (RDF/XML, RDF/JSON, Turtle, N3, TriX, TriG, and others).

If any data space that connects to the Web uses HTTP URIs as identifiers for records, record attributes, and record attribute values, we have Linked Data on the Web as espoused by the Linked Data meme.

Does the model need to be literally “RDF”?

Answer: No.

Does the model need to provide a machine readable mechanism for describing Entities via a constellation of characteristics (properties) that coalesce around the Identifier of the entity being described?

Answer: Yes.

What I described above is simply a “Metadata Model”. A model that enables granular Linkage a on the Web between data items via their de-referencable metadata meshes.

The Linked Data meme is fundamentally about a single HTTP URI as vehicle for three vital things:

1. Unambiguous Identification of an Entity/Data Item/Data Object
2. Accessing the description (metadata) of an Entity/Data Item/Data Object
3. Negotiable representation of Entity/Data Item/Data Object metadata.

If any Web friendly mechanism delivers the above, then it will be compatible with the Linked Data meme.

Evan after all of the above, I still don’t see the need for adding “RDF” and “SPARQL” to the updated Linked Data meme rules. Instead, they should have remained confined to the realm of “suggested implementation options and details”.

Links:

1. http://tr.im/sMoo – Simplified Linked Data meme rules

Kingsley

Clearly your original thoughts where about RDF/XML the format. I discern that with clarity from your comments.

As I indicated in response to Ross, I can bear with current protraction of RDF model and data representation conflation problem; especially, since the version 1.0 of the RDF story is fully to blame for this confusion.

Today though, we do have a clear distinction between the model and its data representational formats (RDF/XML, RDF/JSON, Turtle, N3, TriX, TriG, and others).

If any data space that connects to the Web uses HTTP URIs as identifiers for records, record attributes, and record attribute values, we have Linked Data on the Web as espoused by the Linked Data meme.

Does the model need to be literally “RDF”?

Answer: No.

Does the model need to provide a machine readable mechanism for describing Entities via a constellation of characteristics (properties) that coalesce around the Identifier of the entity being described?

Answer: Yes.

What I described above is simply a “Metadata Model”. A model that enables granular Linkage a on the Web between data items via their de-referencable metadata meshes.

The Linked Data meme is fundamentally about a single HTTP URI as vehicle for three vital things:

1. Unambiguous Identification of an Entity/Data Item/Data Object
2. Accessing the description (metadata) of an Entity/Data Item/Data Object
3. Negotiable representation of Entity/Data Item/Data Object metadata.

If any Web friendly mechanism delivers the above, then it will be compatible with the Linked Data meme.

Evan after all of the above, I still don’t see the need for adding “RDF” and “SPARQL” to the updated Linked Data meme rules. Instead, they should have remained confined to the realm of “suggested implementation options and details”.

Links:

1. http://tr.im/sMoo – Simplified Linked Data meme rules

Kingsley