Copyright © 2008 W3C® (MIT, ERCIM, Keio), All Rights Reserved. W3C liability, trademark and document use rules apply.
Today's web is built predominantly for human consumption. Even as machine-readable data begins to appear on the web, it is typically distributed in a separate file, with a separate format, and very limited correspondence between the human and machine versions. As a result, web browsers can provide only minimal assistance to humans in parsing and processing web data: browsers only see presentation information. We introduce RDFa, which provides a set of XHTML attributes to augment visual data with machine-readable hints. We show how to express simple and more complex datasets using RDFa, and in particular how to turn the existing human-visible text and links into machine-readable data without repeating content.
This document provides only a Primer to RDFa. The normative specification of RDFa can be found in [RDFA-SYNTAX].
This section describes the status of this document at the time of its publication. Other documents may supersede this document. A list of current W3C publications and the latest revision of this technical report can be found in the W3C technical reports index at http://www.w3.org/TR/.
This document is a Working Group Note produced jointly by the W3C Semantic Web Deployment Working Group [SWD-WG] and the W3C XHTML2 Working Group [XHTML2-WG]. This work is part of both the W3C Semantic Web Activity and the HTML Activity. The transition of this document to Working Group Note occurs simultaneously with the transition of the RDFa syntax specification to W3C Recommendation.
This version of the RDFa Primer contains small editorial changes to the previous version as well as a short additional section (4.1) providing pointers to those wishing to create new relationship vocabularies. The changes are detailed in a differences document. The Working Groups have received suggestions that this document be expanded and the Groups may add to it in the future but are not committing to do so.
Comments on this Working Group Note are welcome and may be sent to [email protected]; please include the text "comment" in the subject line. All messages received at this address are viewable in a public archive.
Publication as a Working Group Note does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.
This document was produced groups operating under the 5 February 2004 W3C Patent Policy. W3C maintains a public list of any patent disclosures made in connection with the deliverables of the XHTML 2 group and another public list of any patent disclosures made in connection with the deliverables of the Semantic Web Deployment Working Group; those pages also include instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) must disclose the information in accordance with section 6 of the W3C Patent Policy.
1 Introduction
1.1 HTML vs. XHTML
2 Adding Flavor to XHTML
2.1 Licensing your Work
2.2 Labeling the Title and Author
2.3 Multiple Items per Page
3 Going Deeper
3.1 Contact Information
3.2 Social Network
4 You Said Something about RDF?
4.1 Custom Vocabularies
5 Find Out More
6 Acknowledgments
7 Bibliography
The web is a rich, distributed repository of interconnected information organized primarily for human consumption. On a typical web page, an XHTML author might specify a headline, then a smaller sub-headline, a block of italicized text, a few paragraphs of average-size text, and, finally, a few single-word links. Web browsers will follow these presentation instructions faithfully. However, only the human mind understands that the headline is, in fact, the blog post title, the sub-headline indicates the author, the italicized text is the article's publication date, and the single-word links are categorization labels. The gap between what programs and humans understand is large.
![]() |
What if the browser received information on the meaning of a web page's visual elements? A dinner party announced on a blog could be easily copied to the user's calendar, an author's complete contact information to the user's address book. Users could automatically recall previously browsed articles according to categorization labels (often called tags). A photo copied and pasted from a web site to a school report would carry with it a link back to the photographer, giving her proper credit. When web data meant for humans is augmented with hints meant for computer programs, these programs become significantly more helpful, because they begin to understand the data's structure.
RDFa allows XHTML authors to do just that. Using a few simple XHTML attributes, authors can mark up human-readable data with machine-readable indicators for browsers and other programs to interpret. A web page can include markup for items as simple as the title of an article, or as complex as a user's complete social network.
RDFa benefits from the extensive power of RDF [RDF], the W3C's standard for interoperable machine-readable data. However, readers of this document are not expected to understand RDF. Readers are expected to understand at least a basic level of XHTML.
To date, because XHTML is extensible while HTML is not, RDFa has only been specified for XHTML 1.1. Web publishers are welcome to use RDFa markup inside HTML4: the design of RDFa anticipates this use case, and most RDFa parsers will recognize RDFa attributes in any version of HTML. The authors know of no deployed Web browser that will fail to present an HTML document as intended after adding RDFa markup to the document. However, publishers should be aware that RDFa will not validate in HTML4 at this time. RDFa attributes validate in XHTML, using the XHTML1.1+RDFa DTD.
Consider Alice, a blogger who publishes a mix of professional and personal articles at http://example.com/alice
. We will construct markup examples to illustrate how Alice can use RDFa. The complete markup of these examples can be viewed independently.
In her blog's footer, Alice declares her content to be freely reusable, as long as she receives due credit when her articles are cited. The XHTML includes a link to a Creative Commons [CC] license:
... All content on this site is licensed under <a href="http://creativecommons.org/licenses/by/3.0/"> a Creative Commons License </a>.
A human clearly understands this sentence, in particular the meaning of the link with respect to the current document: it indicates the document's license, the conditions under which the page's contents are distributed. Unfortunately, when Bob visits Alice's blog, his browser sees only a plain link that could just as well point to one of Alice's friends or to her resume. For Bob's browser to understand that this link actually points to the document's licensing terms, Alice needs to add some flavor, some indication of what kind of link this is.
She can add this flavor using the rel
attribute (which we'll write as @rel
so as not to repeat the word "attribute" too often), which defines the relationship between the current page and the linked page. The value of the attribute is license
, an XHTML keyword reserved for just this purpose:
...
All content on this site is licensed under
<a rel="license" href="http://creativecommons.org/licenses/by/3.0/">
a Creative Commons License
</a>.
With this small update, Bob's browser will now understand that this link has a flavor: it indicates the blog's license.
![]() |
Alice is happy that adding XHTML flavor lets Bob find the copyright license on her work quite easily. But what about the article title and author name? Here, instead of marking up a link, Alice wants to augment existing text within the page. The title is a headline, and her name a sub-headline:
<div> <h2>The trouble with Bob</h2> <h3>Alice</h3> ... </div>
To indicate that h2
represents the title of the page, and h3
the author, Alice uses @property
, an attribute introduced by RDFa for the specific purpose of marking up existing text in an XHTML page.
<div xmlns:dc="http://purl.org/dc/elements/1.1/"> <h2 property="dc:title">The trouble with Bob</h2> <h3 property="dc:creator">Alice</h3> ... </div>
Why use dc:creator
and dc:title
, instead of simply creator
and title
? As it turns out, XHTML does not have reserved keywords for those two concepts. Alice could boldly choose to write property="title"
, but how does a program reading this know whether "title" here refers to the title of a work, a job title, or the deed for some real-estate property? And, if every web publisher laid claim to their own short keywords, the world of available properties would become quite messy, a bit like saving every file on a computer's desktop without any directory structure to organize them.
To enforce a modicum of organization, RDFa does not recognize property="title"
. Instead, Alice must indicate a directory somewhere on the web, using simply a URL, from where to import the specific creator
and title
concepts she means to express. Fortunately, the Dublin Core [DC] community has already defined a vocabulary of useful concepts for describing documents, including both creator
and title
, where title
indeed means the title of a work. So, Alice:
xmlns:dc="http://purl.org/dc/elements/1.1/"
, which associates the prefix dc
with the URL http://purl.org/dc/elements/1.1/
, anddc:creator
and dc:title
. These are short-hands for the full URLs http://purl.org/dc/elements/1.1/creator
, and http://purl.org/dc/elements/1.1/title
.
In RDFa, all property names are, in fact, URLs.
![]() |
Alice's blog contains, of course, multiple entries. Sometimes, Alice's sister Eve guest blogs, too. The front page of the blog lists the 10 most recent entries, each with its own title, author, and introductory paragraph. How, then, should Alice mark up the title of each of these entries individually even though they all appear within the same web page? RDFa provides @about
, an attribute for specifying the exact URL to which the contained RDFa markup applies:
<div xmlns:dc="http://purl.org/dc/elements/1.1/"> <div about="/alice/posts/trouble_with_bob"> <h2 property="dc:title">The trouble with Bob</h2> <h3 property="dc:creator">Alice</h3> ... </div> <div about="/alice/posts/jos_barbecue"> <h2 property="dc:title">Jo's Barbecue</h2> <h3 property="dc:creator">Eve</h3> ... </div> ... </div>
We can represent this, once again, as a diagram connecting URLs to properties:
![]() |
Alice can use the same technique to give her friend Bob proper credit when she posts one of his photos:
<div about="/alice/posts/trouble_with_bob"> <h2 property="dc:title">The trouble with Bob</h2> The trouble with Bob is that he takes much better photos than I do: <div about="http://example.com/bob/photos/sunset.jpg"> <img src="http://example.com/bob/photos/sunset.jpg" /> <span property="dc:title">Beautiful Sunset</span> by <span property="dc:creator">Bob</span>. </div> </div>
Notice how the innermost @about
value, http://example.com/bob/photos/sunset.jpg
, "overrides" the outer value /alice/posts/trouble_with_bob
for all markup inside the innermost div
. And, once again, as a diagram that abstractly represents the underlying data of this new portion of markup:
![]() |
In addition, Alice wants to make information about herself (email address, phone number, etc.) easily available to her friends' contact management software. This time, instead of describing the properties of a web page, she's going to describe the properties of a person: herself. To do this, she adds deeper structure, so that she can connect multiple items that themselves have properties.
Alice already has contact information displayed on her blog.
<div> <p> Alice Birpemswick </p> <p> Email: <a href="mailto:[email protected]">[email protected]</a> </p> <p> Phone: <a href="tel:+1-617-555-7332">+1 617.555.7332</a> </p> </div>
The Dublin Core vocabulary does not provide property names for describing contact information, but the Friend-of-a-Friend [FOAF] vocabulary does. In RDFa, it is common and easy to combine different vocabularies in a single page. Alice imports the FOAF vocabulary and declares a foaf:Person
. For this purpose, Alice uses @typeof
, an RDFa attribute that is specifically meant to declare a new data item with a certain type:
<div typeof="foaf:Person" xmlns:foaf="http://xmlns.com/foaf/0.1/">
...
Then, Alice can indicate which content on the page represents her full name, email address, and phone number:
<div typeof="foaf:Person" xmlns:foaf="http://xmlns.com/foaf/0.1/"> <p property="foaf:name"> Alice Birpemswick </p> <p> Email: <a rel="foaf:mbox" href="mailto:[email protected]">[email protected]</a> </p> <p> Phone: <a rel="foaf:phone" href="tel:+1-617-555-7332">+1 617.555.7332</a> </p> </div>
Note how Alice didn't specify @about
like she did when adding blog entry metadata. What is she associating these properties with, then? In fact, the @typeof
on the enclosing div
implicitly sets the subject of the properties marked up within that div
. The name, email address, and phone number are associated with a new node of type foaf:Person
. This node has no URL to identify it, so it is called a blank node.
![]() |
Next, Alice wants to add information about her friends, including at least their names and homepages. Her plain XHTML is:
<div> <ul> <li> <a href="http://example.com/bob/">Bob</a> </li> <li> <a href="http://example.com/eve/">Eve</a> </li> <li> <a href="http://example.com/manu/">Manu</a> </li> </ul> </div>
First, Alice indicates that all of these friends are of type foaf:Person
.
<div xmlns:foaf="http://xmlns.com/foaf/0.1/"> <ul> <li typeof="foaf:Person"> <a href="http://example.com/bob/">Bob</a> </li> <li typeof="foaf:Person"> <a href="http://example.com/eve/">Eve</a> </li> <li typeof="foaf:Person"> <a href="http://example.com/manu/">Manu</a> </li> </ul> </div>
Beyond declaring the type of data we're dealing with, each @typeof
creates a new blank node with its own distinct properties, all without having to provide URL identifiers. Thus, Alice can easily indicate each friend's homepage:
<div xmlns:foaf="http://xmlns.com/foaf/0.1/"> <ul> <li typeof="foaf:Person"> <a rel="foaf:homepage" href="http://example.com/bob/">Bob</a> </li> <li typeof="foaf:Person"> <a rel="foaf:homepage" href="http://example.com/eve/">Eve</a> </li> <li typeof="foaf:Person"> <a rel="foaf:homepage" href="http://example.com/manu/">Manu</a> </li> </ul> </div>
And, of course, each friend's name:
<div xmlns:foaf="http://xmlns.com/foaf/0.1/"> <ul> <li typeof="foaf:Person"> <a property="foaf:name" rel="foaf:homepage" href="http://example.com/bob/">Bob</a> </li> <li typeof="foaf:Person"> <a property="foaf:name" rel="foaf:homepage" href="http://example.com/eve/">Eve</a> </li> <li typeof="foaf:Person"> <a property="foaf:name" rel="foaf:homepage" href="http://example.com/manu/">Manu</a> </li> </ul> </div>
Using @property
, Alice specifies that the linked text ("Bob", "Eve", and "Manu") are, in fact, her friends' names. With @rel
, she indicates that the clickable links are her friends' homepages. Alice is ecstatic that, with so little additional markup, she's able to fully express both a pleasant human-readable page and a machine-readable dataset.
Alice is tired of repeatedly entering information about her friends in each new social networking sites. With RDFa, she can indicate her friendships on her own web page, and let social networking applications read it automatically. So far, Alice has listed three individuals but has not specified her relationship with them; they might be her friends, or they might be her favorite 17th century poets. To indicate that she, in fact, knows them, she uses the FOAF property foaf:knows
:
<div xmlns:foaf="http://xmlns.com/foaf/0.1/" about="#me" rel="foaf:knows">
<ul>
<li typeof="foaf:Person">
<a property="foaf:name" rel="foaf:homepage" href="http://example.com/bob">Bob</a>
</li>
<li typeof="foaf:Person">
<a property="foaf:name" rel="foaf:homepage" href="http://example.com/eve">Eve</a>
</li>
<li typeof="foaf:Person">
<a property="foaf:name" rel="foaf:homepage" href="http://example.com/manu">Manu</a>
</li>
</ul>
</div>
Using rel="foaf:knows"
once is enough to connect Bob, Eve, and Manu to Alice. This is achieved thanks to the RDFa concept of chaining: because the top-level @rel
is without a corresponding @href
, it connects to any contained node, in this case the three nodes defined by @typeof
. (The @about="#me"
is a FOAF/RDF convention: the URL that represents the person Alice is http://example.com/alice#me
. It should not be confused with Alice's homepage, http://example.com/alice
. You are what you eat, but you are far more than just your homepage.)
![]() |
RDF, the Resource Description Framework, is exactly the abstract data representation we've drawn out as graphs in the above examples. Each arrow in the graph is represented as a subject-predicate-object triple: the subject is the node at the start of the arrow, the predicate is the arrow itself, and the object is the node or literal at the end of the arrow. An RDF dataset is often called an "RDF graph", and it is typically stored in what is often called a "Triple Store."
Consider the first example graph:
![]() |
The two RDF triples for this graph are written, using the Notation3 syntax [N3], as follows:
<http://www.example.com/alice/posts/trouble_with_bob> <http://purl.org/dc/elements/1.1/title> "The Trouble with Bob"; <http://purl.org/dc/elements/1.1/creator> "Alice" .
Also, the TYPE
arrows we drew are no different from other arrows, only their label is actually a core RDF property, rdf:type
, where the rdf
namespace is <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
. The contact information example from above should thus be diagrammed as:
![]() |
The point of RDF is to provide a universal language for expressing data. A unit of data can have any number of fields, and field names are URLs which can be reused by any publisher, much like any web publisher can link to any web page, even ones they did not create themselves. Given data, in the form of RDF triples, collected from various locations, and using the RDF query language SPARQL [SPARQL], one can search for "friends of Alice's who created items whose title contains the word 'Bob'," whether those items are blog posts, videos, calendar events, or other data types we haven't thought of yet.
RDF is an abstract, machine-readable data representation meant to maximize the reuse of vocabularies. RDFa is a way to express RDF data within XHTML, by reusing the existing human-readable data.
As Alice marks up her page with RDFa, she may discover the need to express data, e.g. her favorite photos, that is not covered by existing vocabularies like Dublin Core or FOAF. Since RDFa is simply a representation of RDF, the RDF schema mechanism that enables RDF extensibility is the same that enables RDFa extensibility. Once an RDF vocabulary created, it can be used in RDFa markup just like existing vocabularies.
The instructions on how to create an RDF schema are available in Section 5 of the RDF Primer [RDF-SCHEMA-PRIMER]. At a high level, the creation of an RDF schema for RDFa involves:
http://example.com/photos/vocab#
.Photo
and Camera
, as well as the property takenWith
that relates a photo to the camera with which it was taken.xmlns:photo="http://example.com/photos/vocab#"
, and typeof="photo:Camera"
.It is worth noting that anyone who can publish a document on the Web can publish an RDF vocabulary and thus define new data fields they may wish to express. RDF and RDFa allow fully distributed extensibility of vocabularies.
More examples, links to tools, and information on how to get involved can be found on the the RDFa Wiki.
This document is the work of the RDF-in-HTML Task Force, including (in alphabetical order) Ben Adida, Mark Birbeck, Jeremy Carroll, Michael Hausenblas, Shane McCarron, Steven Pemberton, Manu Sporny, Ralph Swick, and Elias Torres. This work would not have been possible without the help of the Semantic Deployment Working Group and its previous incarnation, the Semantic Web Deployment and Best Practices Working Group, in particular chairs Tom Baker and Guus Schreiber (and prior chair David Wood), the XHTML2 Working Group, Eric Miller, previous head of the Semantic Web Activity, and Ivan Herman, current head of the Semantic Web Activity. Earlier versions of this document were officially reviewed by Gary Ng and David Booth, and more recent versions by Diego Berrueta and Ed Summers, all of whom provided insightful comments that significantly improved the work. Bob DuCharme also reviewed the work and provided useful commentary.