This article describes best practices for marking works licensed with a Creative Commons license or dedicated to the public domain with both human readable and machine readable notices, on the web and otherwise.

Although the importance of human-visible attribution and notice is stressed, this primary focus of this article is machine-readable metadata. Best practice for presentation of human-visible attribution and notice in various formats (apart from colocation with machine-readable equivalents) is currently beyond the scope of this article and will be presented elsewhere.

Principles

• Copyright notice should be published on the web. Offline or peer-distributed works should refer to notice published on the web. A web notice allows content to be found via web search, content owners to be found by viewers of untethered works, may be trusted to the extent a web page may be trusted, and allows for further annotation by the owner and others.

• Copyright notice should be visible and unambiguous to humans and computers, especially the former. Human and computer visible notices should be one and the same.

• Copyright notice best practices should be general rather than applicable only to works offered under one of the Creative Commons licenses.

Web

A web page is the preferred venue for published copyright notices. Notices published elsewhere should refer to an equivalent notice published on the web.

       This text is licensed to the public under the <a rel="license"
        href="http://creativecommons.org/licenses/by-sa/2.5/">Creative
        Commons Attribution 2.5 License</a>.

That HTML snippet says that the current document is licensed under CC BY 2.5. The current document is the default subject, rel="license" sets the predicate or verb, the URL in the href value sets the object.

For detailed background on the web metadata model and syntax being used, see RDFa.

This use follows the aforementioned principles:

• On the web (obviously, so long as the HTML page in question is published to a website).

• The metadata is visible -- it is colocated with an actionable link and may further be styled with CSS for human consumption.

• It is general -- the model used can accomodate any sort of statement about anything that has a URI and the license predicate can take any license that has a URI as its object, not only a Creative Commons license.

Now we will see how to use the model to further annotate works with licensing-relevant metadata.

Attribution

A licensor can specify (see 4(b) of http://creativecommons.org/licenses/by/2.5/legalcode):

1. name of author or designated entity
2. title of work
3. url associated with work referencing copyright
4. identification of derivative use

Ideally all of these would be available as machine-readable metadata. I think the triples we want (assuming the work in question is the current document) are:

       1. <> cc:attributionName "James Roberts" .
       2. <> dc:title "Compact Representation of Blank Pages" .
       3. <> cc:attributionURL <http://example.org/crobp.html> .
       4. <> cc:derivativeDescription "translation of 'Paginas Blancos Si!'" .

Why the new properties?

1. dcterms:rightsHolder and dc:creator don't say which should be attributed -- or if someone else should be.
2. not new, dc:title works just fine.
3. often this will be the current URL, but need some way to specify that it should be cited.
4. dc:source refers to the source work, not how source work was reused.

Fortunately RDFa allows us to annotate human readable notices and include both the custom properties needed for attribution as well as additional useful properties, e.g.

       <a rel="cc:attributionURL" href="http://example.org/crobp.html"
       property="dc:title">Compact Representation of Blank Pages</a> by
       <a rel="dc:creator" href="http://example.org/jr.html"
       property="cc:attributionName">James Roberts</a>, a <a
       rel="dc:source" href="http://example.org/bps.html"
       property="cc:derivativeDescription">translation of 'Paginas
       Blancos Si!'</a>, is licensed to the public under the <a
       rel="license"
       href="http://creativecommons.org/licenses/by/2.5/">Creative
       Commons Attribution 2.5 License</a>.

This produces the four triples above plus these:

       5. <> dc:creator <http://example.org/jr.html> .
       6. <> dc:source <http://example.org/bps.html> .
       7. <> license <http://creativecommons.org/licenses/by/2.5/> .

Included Objects

Objects included in web pages (images) should be annotated as above within their "host" web pages. Ideally all included objects should also embed object format native notice and metadata as described below, but this is only crucial for objects where the publisher intends for or is concerned about distribution outside the context of the "host" web page.

Metadata about include objects that have their own URIs should be qualified with about, as rel="license" without an about attribute makes a statement about the current document (which is the "host" web page). Example:

       Photo licensed under <a about="http://example.com/some-image.png"
        rel="license" href="http://creativecommons.org/licenses/by/2.5/">cc by 2.5</a>.

See the RDFa Primer on beyond the current document for additional examples.

More Rights

A page may provide any sort of metadata for itself and its included components. The following example has some obviously useful statements:

       <span rel="dc:type" href="http://purl.org/dc/dcmitype/Sound"/>Audio</span> <a rel="license"
        href="http://creativecommons.org/licenses/by-nc/2.5/">(cc)</a> <span property="dc:date">2006</span>.
        See <a rel="cc:morePermissions" href="http://johnsmith.com/store">my store</a> to obtain permissions not
        granted by the CC license, signed CDs, and concert tickets.

In addition to the familiar license statement and dc:type from Dublin Core we have cc:morePermissions. The intention of cc:moreRights is to point to a URL at which one can discover permissions outside the scope of those granted to the public by the licensor via a CC license.

A prominent use case is Magnatune's commercial licensing system. Example:

• http://magnatune.com/artists/albums/anup-embrace/ "Embrace" by Anup
• cc:morePermissions for this work would point to https://magnatune.com/artists/license/?artist=Anup&album=Embrace&genre=World where one can obtain commercial licensing for advertising, film, video games, etc.

Namespaces

In the example snippets above we have used the cc: and dc: namespaces without definition. To define a namespace for subsequent use include an xmlns attribute on the element in which the namespace is used on any parent element. Example:

       <span xmlns:dc="http://purl.org/dc/elements/1.1/" property="dc:title">My Book</span>

Interoperability with other web-based metadata formats

RDF/XML

RDF/XML is another RDF serialization that can be used to make any statement that can be made with RDFa. RDF/XML may be <link>ed in the <head> of a web page or (deprecated CC recommendation) included in HTML comments. RDF/XML is perfectly interoperable with RDFa, but the latter is preferred as it is colocated with human-visible content.

Microformats

RelLicense is both an elemental microformat (specifically rel license) and an RDFa statement. Compound Microformats should be interoperable via GRDDL or hGRDDL.

Use cases for licensing and related metadata as microformats may be found at on the microformats wiki; they remain hypothetical as microformats, but generalize to use cases for Creative Commons web metadata.

Feeds

RSS1.0, RSS2.0, and Atom 1.0 syndication formats each may include license annotations. These should also be reflected on web pages referenced by feed items.

Non-Web

Objects not native to the web should adhere to the following principles regarding license notice and metadata.

• To the extent possible copyright notice should be visible to users of the licensed object.

• The licensed object should contain a web reference to a web page that provides an equivalent licensing notice for the object in question. This makes a non-web object's licesning status as certain as that of a web page's status -- if the web reference does not agree with the non-web object's license notice, ignore the latter.

• To the extent possible license notice provided by non-web objects should use mechanisms and conventions in common use for the format of the non-web object in question.

We will now apply these principles to a variety of content formats, after describing two building blocks for Non-Web metadata -- web metadata and XMP.

Web Statement

The web page that provides licensing notice equivalent to any included in the object itself may be only human readable. Ideally the web page will publish metadata that is explicitly associated with the object in question. To do this the object must be identified, ideally with a content-derived identifier, such that a client may verify that the web page metadata concerns precisely the object in question. An example of a content-derived identifier is a content hash.

Following is an example of RDFa describing a resource indentified by a hash:

   <span about="urn:sha1:DATAG7ENBVHFNQPM4W626VDVK25RYECI">
    'Good Dream' is licensed under
    <a rel="license" href="http://creativecommons.org/licenses/by/2.5/">CC BY</a>.
    </span>

Produces the following triple:

       <urn:sha1:DATAG7ENBVHFNQPM4W626VDVK25RYECI> license <http://creativecommons.org/licenses/by/2.5/> .

XMP

XMP has the broadest support of any embedded metadata format (perhaps it is the only such format with anything approaching broad support) across many different media formats. With the exception meida formats where a workable embedded metadata format is already ubiquituous, Creative Commons recommends adopting XMP as an embedded metadata standard and use of the following two fields in particluar:

• Web reference: value of xapRights:WebStatement
• License: value of cc:license

Audio

Human visible metadata is not possible, and audio notice is not acceptable in most music contexts. With the exception of MP3 and OGG below, XMP embedded metadata is recommended.

MP3

MP3 is the primary exception to the XMP recommendation above, as ID3 is widely supported. The following two fields are recommended:

• Web reference: value of WOAF ("official audio file" URL)
• License: value of WCOP (copyright URL)

OGG

"Vorbis Comments" are widely supported for OGG files. The following two fields are recommended:

• Web reference: value of CONTACT
• License: value of LICENSE

Note that OGG is a container format that also supports video.

Still Images

Although possible, visible attribution and copyright notice generally either does not fit in an image due to size or aesthetic limitations. Even where possible human-visible notice will not be colocated with machine-readable metadata (for bitmap formats currently in use).

For embedded metadata XMP is recommended.

Video

Visible attribution and copyright notice is generally provided in video frames (e.g., credit roll) or overlays. An example of such an overlay is available here: http://wiki.creativecommons.org/Marking_work#Video JordanN 11:56, 5 March 2007 (UTC)

Preferably relevant regions of the video will be clickable or otherwise interactively linked with web pages.

There are many video formats, each with its own ill-supported embedded metadata specification (if any). Creative Commons recommends adopting XMP across video formats, with the possible exception of OGG (see audio above).

Document formats

"Document" formats meaning word processor, presentation and spreadsheet document files or their output formats such as PDF.

All of these formats are intended for human consumption and all modern versions support web links, so visibility is easy -- the license and web reference should be noted in thee document text wherever common use dictates a copyright notice would appear. Looks like we just took care of the other principles as well.

Many document formats support some form of embedded metadata that can be accessed, e.g., by selecting File|Document Properties from a menu. Where such metadata includes licensing or copyright-relevant fields, these should be populated with first priority to the web reference and second priority to the URL of the license the document in question is under.

Ideally these metadata fields should be used to directly populate the human-visible parts of the document, e.g., via a document field, as is typically used to automatically place page numbers and document titles throughout such documents. This gives us the best of both worlds -- any software that indexes embedded metadata can easily find license information, and that information is perfectly reflected in notices humans see.

What follows is a very incomplete accounting of how these guidelines may be used in various popular office document formats.

PDF

Visible notice as above. Use XMP embedded metadata.

Microsoft Office

Visible notice as above. Metadata TBD.

OpenOffice.org

Visible notice as above. Metadata TBD.