Syntax Comparison

This conclusion is obviously a bit biased, but there is certainly room for more than RDF syntax.

For augmenting HTML documents with small amounts of additional metadata, eRDF and RDFa are good choices.

If compatibility with existing XHTML1.x DTDs is a concern, then eRDF is best - if its subset of RDF is sufficient;

If you are only planning on reading the entire file into an RDF model, then Turtle or RDF/XML are good choices.

For consumers not using an off-the-shelf RDF API, then Turtle is easier to write a parser for, and is more intuitive.

RXR, TriX, and N-TRIPLES are easy to parse to triples, but their verbosity and poor human-readability may make other formats a better choice.

treetriples is a good choice, although its special support for XML Literals requires support for exclusive canonicalization, which makes parsing to triples slightly harder than RXR.

For accessing and querying the RDF directly with tree-oriented XML technologies, treetriples is ideal.

RXR and TriX are alternatives, but they are verbose, and have poor human readability. Also, because the triples structure is flat and ungrouped, they are less efficient than treetriples because the entire document needs to be searched for each triple match, whereas treetriples can more efficiently select several triples with the same subject, because only a single branch of the document needs to be searched.

A quick summary of the strengths and weaknesses of various formats

Format	RDF library IO	Human IO	Hackable IO	XML tools
RDF/XML	5	4	0	0
N-Triples	4	1	5	0
TriX / RXR	1	1	4	3
Turtle	3 (read as N3)	5	4	0
RDFa / eRDF	2 (GRDDL)	2	3	1
treetriples	2 (GRDDL)	3	4	5

RDF/XML

[Link to Specification]

Advantages

Blessed by the RDF Rec. Most applications offer support for RDF/XML, due to it being the only format specified by the RDF Recs, regardless of its technical problems.
Very concise. Namespaces as shortcuts for URI prefixes; shortcuts for rdf:type; shortcuts for collections, containers, and reification; using element names to carry content; and defaulting of subjects and (subject, predicate) pairs - all lead to an extremely concise syntax, and the reduction in cruft makes the format more human readable too.
Very human-readable (assuming that it has been human generated, or generated by a careful serializer, and that the data is fairly regular). The striped syntax makes RDF look like a typical XML syntax for the common case of hierarchical data with few subjects and many properties.

Disadvantages

Hard to implement. Implementing an RDF/XML parser is very dificult. It isn’t unusual to find bugs in the RDF/XML parsers of even mature libraries, because RDF/XML relies on tricky technologies such as Namespaces, xml:base, and exclusive canonicalization.
Impossible to process with standard XML tools. To some extent the formats superficial resemblance to vanilla XML does no favours for the adoption of RDF by raising expectations and failing to meet them.
Not general purpose. Some RDF graphs cannot be serialised to RDF/XML, and other reasonable constructs cannot use the shorthand forms for lists and collections.

Example

<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
	 xmlns:ex="http://www.example.com/vocab#"
	 xmlns:html="http://www.w3.org/1999/xhtml">

  <rdf:Description rdf:about="http://www.example.com/thing/1/">
    <ex:plain>Plain Literal</ex:plain>
    <ex:lang xml:lang="en">English Literal</ex:lang>
    <ex:bag>
      <rdf:Bag>
	<rdf:li>1</rdf:li>
	<rdf:li>2</rdf:li>
	<rdf:li>3</rdf:li>
      </rdf:Bag>
    </ex:bag>
    <ex:list rdf:parseType="Collection">
      <rdf:Description rdf:about="http://www.example.com/thing/2/" />
      <rdf:Description rdf:about="http://www.example.com/thing/3/" />
    </ex:list>
    <ex:xml rdf:parseType="Literal">
      Below is:
      <html:div>Some <html:b>mixed</html:b> content.</html:div>
    </ex:xml>
    <ex:uri rdf:resource="http://www.example.com/thing/4/" />
    <ex:bnode rdf:nodeID="b2" />
  </rdf:Description>

  <ex:Type rdf:nodeID="b1">
    <ex:number rdf:datatype="http://www.w3.org/2001/XMLSchema#int">7</ex:number>
  </ex:Type>

</rdf:RDF>

N-Triples

N-Triples isn’t intended to be a general purpose interchange format (it was designed as an alternative syntax for RDF/XML test cases), but it is useful nonetheless, and its extreme simplicity is a useful reference point to some of the other syntaxes listed here.

[Link to Specification]

Advantages

This simple specification is used by the RDF test cases document.
Extremely simple to implement serializers and parsers.
Effective in certain applications where verbosity and readability aren’t a concern, such as streaming RDF between two processes as a form of IPC.

Disadvantages

Very verbose. The format has no abbreviation features at all.
Poor human-readability. All of the verbosity isn’t easy on the eyes.
Not XML, so the format can’t make use of XML tools such as XSLT.

Example

<http://www.example.com/thing/1/> <http://www.example.com/vocab#bnode> _:b1 .
<http://www.example.com/thing/1/> <http://www.example.com/vocab#uri> <http://www.example.com/thing/4/> .
<http://www.example.com/thing/1/> <http://www.example.com/vocab#xml> "\n      Below is:\n      <html:div xmlns:html=\"http://www.w3.org/1999/xhtml\">Some <html:b>mixed</html:b> content.</html:div>\n    "^^<http://www.w3.org/1999/02/22-rdf-syntax-ns#XMLLiteral> .
<http://www.example.com/thing/1/> <http://www.example.com/vocab#list> _:b2 .
<http://www.example.com/thing/1/> <http://www.example.com/vocab#bag> _:b3 .
<http://www.example.com/thing/1/> <http://www.example.com/vocab#lang> "English Literal"@en .
<http://www.example.com/thing/1/> <http://www.example.com/vocab#plain> "Plain Literal" .
_:b4 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> <http://www.w3.org/1999/02/22-rdf-syntax-ns#nil> .
_:b4 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://www.example.com/thing/3/> .
_:b1 <http://www.example.com/vocab#number> "7"^^<http://www.w3.org/2001/XMLSchema#int> .
_:b1 <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://www.example.com/vocab#Type> .
_:b3 <http://www.w3.org/1999/02/22-rdf-syntax-ns#_3> "3" .
_:b3 <http://www.w3.org/1999/02/22-rdf-syntax-ns#_2> "2" .
_:b3 <http://www.w3.org/1999/02/22-rdf-syntax-ns#_1> "1" .
_:b3 <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://www.w3.org/1999/02/22-rdf-syntax-ns#Bag> .
_:b2 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> _:b4 .
_:b2 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://www.example.com/thing/2/> .

TriX

TriX is an XML syntax, that adopts a flat triples style syntax like N-Triples, but it also allows XSLT to be used to provide syntactic extensions that render down to the un-extended TriX format.

[Link to Specification]

Advantages

Due to the regular format it is possible to use XML technologies such as XSLT to process TriX.
TriX has explicit support for named graphs. I haven’t seen named graphs widely used in serialised RDF though.
TriX supports an extensible syntax via XSLT.

Disadvantages

The un-extended syntax is very verbose.
The lack of structure in the un-extended syntax gives it poor human-readability.
Although it is possible to use XSLT to transform TriX documents, it is inefficient because each template match must query the entire document, due to the lack of hierarchy.
Using the xml-stylesheet PI to specify the extension transformation overloads the intended purpose of the PI as a transformation targeted at human presentation. It would be better to use an attribute to specify the stylesheet so that extension pre-processing, and human presentation post-processing could be used independently.

Example

<graphset xmlns="http://jena.sourceforge.net/TriX/" >
  <graph>
    <triple>
      <id>_:b1</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#type</uri>
      <uri>http://www.example.com/vocab#Type</uri>
    </triple>
    <triple>
      <id>_:b1</id>
      <uri>http://www.example.com/vocab#number</uri>
      <typedLiteral datatype="http://www.w3.org/2001/XMLSchema#int">7</typedLiteral>
    </triple>
    <triple>
      <id>_:b2</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#rest</uri>
      <id>_:b3</id>
    </triple>
    <triple>
      <id>_:b2</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#first</uri>
      <uri>http://www.example.com/thing/2/</uri>
    </triple>
    <triple>
      <id>_:b4</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#type</uri>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#Bag</uri>
    </triple>
    <triple>
      <id>_:b4</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#_1</uri>
      <plainLiteral>1</plainLiteral>
    </triple>
    <triple>
      <id>_:b4</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#_2</uri>
      <plainLiteral>2</plainLiteral>
    </triple>
    <triple>
      <id>_:b4</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#_3</uri>
      <plainLiteral>3</plainLiteral>
    </triple>
    <triple>
      <id>_:b3</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#rest</uri>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#nil</uri>
    </triple>
    <triple>
      <id>_:b3</id>
      <uri>http://www.w3.org/1999/02/22-rdf-syntax-ns#first</uri>
      <uri>http://www.example.com/thing/3/</uri>
    </triple>
    <triple>
      <uri>http://www.example.com/thing/1/</uri>
      <uri>http://www.example.com/vocab#bnode</uri>
      <id>_:b1</id>
    </triple>
    <triple>
      <uri>http://www.example.com/thing/1/</uri>
      <uri>http://www.example.com/vocab#list</uri>
      <id>_:b2</id>
    </triple>
    <triple>
      <uri>http://www.example.com/thing/1/</uri>
      <uri>http://www.example.com/vocab#bag</uri>
      <id>_:b4</id>
    </triple>
    <triple>
      <uri>http://www.example.com/thing/1/</uri>
      <uri>http://www.example.com/vocab#lang</uri>
      <plainLiteral xml:lang="en">English Literal</plainLiteral>
    </triple>
    <triple>
      <uri>http://www.example.com/thing/1/</uri>
      <uri>http://www.example.com/vocab#plain</uri>
      <plainLiteral>Plain Literal</plainLiteral>
    </triple>
    <triple>
      <uri>http://www.example.com/thing/1/</uri>
      <uri>http://www.example.com/vocab#xml</uri>
      <typedLiteral datatype="http://www.w3.org/1999/02/22-rdf-syntax-ns#XMLLiteral">
	Below is:
	&lt;html:div xmlns:html="http://www.w3.org/1999/xhtml"&gt;Some &lt;html:b&gt;mixed&lt;/html:b&gt; content.&lt;/html:div&gt;
      </typedLiteral>
    </triple>
    <triple>
      <uri>http://www.example.com/thing/1/</uri>
      <uri>http://www.example.com/vocab#uri</uri>
      <uri>http://www.example.com/thing/4/</uri>
    </triple>
  </graph>
</graphset>

RXR

RXR is similar to TriX, in that it is a simple, triple-based XML serialization of RDF. It removes syntactic extensibility and named-graphs, but adds a concise syntax for collections.

[Link to Specification]

Advantages

Due to the regular format it is possible to use XML technologies such as XSLT to process RXR.
Simpler than TriX
More concise than un-extended TriX, due to the collections abbreviation.

Disadvantages

Still rather verbose.
The verbosity and lack of structure also lead to poor human readability.
Although it is possible to use XSLT to transform RXR documents, it is inefficient because each template match must query the entire document, due to the lack of hierarchy.

Example

<graph xmlns="http://ilrt.org/discovery/2004/03/rxr/">
  <triple>
    <subject blank="_:b1" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#type" />
    <object uri="http://www.example.com/vocab#Type" />
  </triple>
  <triple>
    <subject blank="_:b1" />
    <predicate uri="http://www.example.com/vocab#number" />
    <object datatype="http://www.w3.org/2001/XMLSchema#int">7</object>
  </triple>
  <triple>
    <subject blank="_:b2" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#rest" />
    <object blank="_:b4" />
  </triple>
  <triple>
    <subject blank="_:b2" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#first" />
    <object uri="http://www.example.com/thing/2/" />
  </triple>
  <triple>
    <subject blank="_:b3" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#type" />
    <object uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#Bag" />
  </triple>
  <triple>
    <subject blank="_:b3" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#_1" />
    <object>1</object>
  </triple>
  <triple>
    <subject blank="_:b3" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#_2" />
    <object>2</object>
  </triple>
  <triple>
    <subject blank="_:b3" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#_3" />
    <object>3</object>
  </triple>
  <triple>
    <subject blank="_:b4" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#rest" />
    <object uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#nil" />
  </triple>
  <triple>
    <subject blank="_:b4" />
    <predicate uri="http://www.w3.org/1999/02/22-rdf-syntax-ns#first" />
    <object uri="http://www.example.com/thing/3/" />
  </triple>
  <triple>
    <subject uri="http://www.example.com/thing/1/" />
    <predicate uri="http://www.example.com/vocab#bnode" />
    <object blank="_:b1" />
  </triple>
  <triple>
    <subject uri="http://www.example.com/thing/1/" />
    <predicate uri="http://www.example.com/vocab#list" />
    <object blank="_:b2" />
  </triple>
  <triple>
    <subject uri="http://www.example.com/thing/1/" />
    <predicate uri="http://www.example.com/vocab#bag" />
    <object blank="_:b3" />
  </triple>
  <triple>
    <subject uri="http://www.example.com/thing/1/" />
    <predicate uri="http://www.example.com/vocab#lang" />
    <object xml:lang="en">English Literal</object>
  </triple>
  <triple>
    <subject uri="http://www.example.com/thing/1/" />
    <predicate uri="http://www.example.com/vocab#plain" />
    <object>Plain Literal</object>
  </triple>
  <triple>
    <subject uri="http://www.example.com/thing/1/" />
    <predicate uri="http://www.example.com/vocab#xml" />
    <object datatype="http://www.w3.org/1999/02/22-rdf-syntax-ns#XMLLiteral">
      Below is:
      &lt;html:div xmlns:html="http://www.w3.org/1999/xhtml"&gt;Some &lt;html:b&gt;mixed&lt;/html:b&gt; content.&lt;/html:div&gt;
    </object>
  </triple>
  <triple>
    <subject uri="http://www.example.com/thing/1/" />
    <predicate uri="http://www.example.com/vocab#uri" />
    <object uri="http://www.example.com/thing/4/" />
  </triple>
</graph>

RPV

RPV looks superficially similar to treetriples because it is an XML syntax with concise element names.

[Link to Specification]

Advantages

Concise reification syntax.
xml:base-style URI abbreviation via rBase, pBase, and vBase
Abbreviation of multiple properties of the same resource.
Keeps crufty implementation details of the XML serialization to a minimum in the document by using short element names.

Disadvantages

The specification may be incomplete?
- is it possible to have bNodes as the object of properties?
- what is the purpose of the vText attribute?
- support for language-tagged literals or typed literals?
Encouraging URIs to be abbreviated by using an xml:base style syntax to arbitrarily split them, makes the syntax difficult to process with XML tools such as XSLT.

Turtle

Turtle is extremely concise, extremely human-readable, and extremely machine-readable. The only disadvantage is that it isn’t XML.

[Link to Specification]

Advantages

Very concise. QNames as shortcuts for URIs; shortcuts for rdf:type; shortcuts for collections; and defaulting of subjects and (subject, predicate) pairs - all lead to an extremely concise syntax, and the reduction in cruft makes the format more human readable too.
The hierarchical grouping gives the syntaxes a natural feel which makes it more human readable.
It is fairly easy to implement a parser and serializer.

Disadvantages

Not XML, so cannot use XML technologies such as XSLT.

Example

@prefix :        <#> .
@prefix ex:      <http://www.example.com/vocab#> .
@prefix rdf:     <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .

<http://www.example.com/thing/1/>
      ex:bag  [ rdf:type rdf:Bag ;
                rdf:_1  "1" ;
                rdf:_2  "2" ;
                rdf:_3  "3"
              ] ;
      ex:bnode
              [ rdf:type ex:Type ;
                ex:number "7"^^<http://www.w3.org/2001/XMLSchema#int>
              ] ;
      ex:lang "English Literal"@en ;
      ex:list (<http://www.example.com/thing/2/> <http://www.example.com/thing/3/>) ;
      ex:plain "Plain Literal" ;
      ex:uri  <http://www.example.com/thing/4/> ;
      ex:xml  "\n      Below is:\n      <html:div xmlns:html=\"http://www.w3.org/1999/xhtml\">Some <html:b>mixed</html:b> content.</html:div>\n    "^^rdf:XMLLiteral .

RDFa

RDFa is an extension of XHTML that allows RDF triples to be embedded in an HTML document. The work is currently still in progress.

[Link to Specification]

Advantages

Co-locating metadata with human-oriented documentation makes it easier to maintain the data.
Makes is possible to use RDF with successful HTML oriented web applications (eg: CMSs, myspace).

Disadvantages

RDFa does not validate according to the standard XHTML DTDs because it adds additional elements and attributes. This isn’t such a serious architectural concern, but does cause practical issues such as incompatibilities with authoring tools and the W3C online HTML validator.
CURIES rely on the context of in-scope namespace declarations to interpret QNames. Some would consider this to be an anti-pattern.
RDFa may be difficult for RDF applications to write and edit. The benefit of RDFa is that it allows metadata to be co-located with human readable text; this tends to be done by a human, a machine may have difficulty knowing the best place to write all triples in a graph without some additional data-specific knowledge.

eRDF

[Link to Specification]

Advantages

Provides a natural way of embedding RDF in existing HTML documents.
Co-locating metadata with human-oriented documentation makes it easier to maintain the data.
Makes is possible to use RDF with successful HTML oriented web applications (eg: CMSs, myspace).
Validates according to standard XHTML DTDs.
Doesn’t rely on QNames-in-content.

Disadvantages

Only supports a fairly small subset of RDF.
eRDF may be difficult for RDF applications to write and edit. The benefit of eRDF is that it allows metadata to be co-located with human readable text; this tends to be done by a human, a machine may have difficulty knowing the best place to write all triples in a graph without some additional data-specific knowledge.

Custom GRDDL + XHTML syntaxes

[Link to Specification]

Advantages

A custom transform can potentially reduce data-duplication and verbosity further than eRDF or RDFa, making maintenance even simpler.
Would normally validate according to standard XHTML DTDs.

Disadvantages

Likely to only support a small subset of RDF.

treetriples

The proposal is like an XML reformulation of Turtle, in that it uses nesting to abbreviate subjects with multiple properties and (subject, property) pairs with multiple values.

Like RPV it uses short tag names to reduce verbosity and improve readability.

Unlike most other RDF syntaxes, treetriples has strict requirements of how to serialized documents. In order to make documents easier to process,

[Link to Specification]

Advantages

The syntax can be easily processed using XSLT, allowing HTML, XML, and other formats to be easily generated from the document.
XML Literals can be processed with XSLT, just like the surrounding syntax. This is better than relegating XML literals to be opaque text blobs that would require separate parsing, as they would be in many RDF model APIs.

Disadvantages

The strict requirements make the syntax more difficult to generate than other syntaxes, especially from XSLT.
XML Literals need to be converted to exclusive canonical XML if they need to be converted to string literals by the consumer, this adds some complexity.

Syntax Comparison

Conclusion

A quick summary of the strengths and weaknesses of various formats

RDF/XML

Advantages

Disadvantages

Example

N-Triples

Advantages

Disadvantages

Example

TriX

Advantages

Disadvantages

Example

RXR

Advantages

Disadvantages

Example

RPV

Advantages

Disadvantages

Turtle

Advantages

Disadvantages

Example

RDFa

Advantages

Disadvantages

eRDF

Advantages

Disadvantages

Custom GRDDL + XHTML syntaxes

Advantages

Disadvantages

treetriples

Advantages

Disadvantages

Example