|
LinkingImplementationProposals
This page outlines the various proposals for EPUB Linking & Identification. Problem ScopeThe following implementation proposals attempt to solve a problem (linking) that is effectively 2-fold:
One design goal is to enable the encoding of both types of information within a single format, based (most likely) on the Uniform Resource Identifier syntax (URI - RFC3986). However, for the sake of clarity, it would be desirable to clearly distinguish Part A (identification of an EPUB publication) and Part B (identification of content within an EPUB publication). In the future, this dichotomy would allow the design scope for more sophisticated linking use-cases (such as annotating documents containing video/audio clips) to be de-facto restricted to Part B only, leaving Part A untouched. In this version of EPUB3, the emphasis is put on the simple "fragment identifier" linking mechanism, for referencing XML/XHTML content in a relatively abstract manner. URI Links ProposalIntroductionHistorically each EPUB document has been identified using a “unique-identifier” property which references a meta element containing the unique identifier by id. Although this identifier has in theory been unique, there has been no prescribed method for insuring its uniqueness across the universe of publishers. In addition, it has been unclear when to change a publication’s identifier. Some publishers change the identifier each time a new version is created, while others change the identifier only when a new “work” is created. As EPUBs have increased in prominence and longevity, it has become important to introduce a permanent unambiguous way of referring to an EPUB both offline and online. This proposal seeks to describe an approach to solving these problems using URIs. URI SyntaxEach EPUB will be identified by a URI that is constructed from the following components: schemeThe URI scheme such as “http” or “ftp”. The scheme may be sued as a protocol to access the EPUB in the event that the URI happens to be a URL but there is no mandate that this be the case. authorityThe authority under which the URI has been created. The authority must be DNS registered to insure that URIs are globally unique. Example authorities include: “epubs.idpf.org” or “randomhouse.com”. pathThe publisher controlled path of the EPUB. Publishers may choose paths in any way that is convenient to them. For example if one wished to organize content by language, one might have paths like: “/epubs/english/history” and “/epubs/japanese/cooking”. unique-identifierThe unique identifier string identified in the EPUB’s OPF file. The definition of the unique-identifier for EPUB3 will be augmented to specify the unique-identifier of a publication will only be changed when a new “work” is created. versionThe version number specified using the new EPUB3 “opf:content-version” property specified in the EPUB. Version numbers may be specified with a varying precision. For example, if an EPUB has a version of “1.2.511” it should be considered as matching URIs that specify a version number of “1.2” or “1” or which omit the “version” segment of the URI entirely. A reference that fails to specify a more granular version number should receive the highest version available in response. fileThe name of a file within the manifest of an EPUB. This URI segment is entirely optional and in general will be omitted. It will only be present when one wishes to refer to a specific file within an EPUB to provide a deep linking capability. idThe id portion of the URI will only be specified when a link refers to a specific xml element within an EPUB content file. In general the id field will be omitted. If an id is specified, a file must be specified. Canonical URI syntaxFollowing is the canonical form a URI that may be used reference an EPUB. The segments contained within “{}” brackets may be omitted. scheme:://authority/path/unique-identifier{/version}{{/file}#id}Example URIs http://epub.idpf.org/us/0741021137
http://epub.idpf.org/us/0741021137/1.2.511
epub://mydomain.com/alice_in_wonderlandURI ProcessingAll user agents that handle EPUB links are required to handle them in such a way that version numbers degrade gracefully as described above under “version”. This would mean that if one were to process EPUB URI requests over the web, it would be necessary to return success from multiple locations depending on the specificity of the link provided. Handling EPUB2 PublicationsEPUB2 publications do contain unique identifiers, but they do not contain version numbers. As a result, EPUB2 publications must only be referred to using a URI that omits the version segment. Because EPUB2 publications do not contain a scheme, authority or path, a default for these fields must be provided. Whenever there is no scheme authority or path specified, the default, for processing purposes will be assumed to be “http://idpf.org/epub”. Open Issues
PURL-based EPUB IdentifiersSummaryA PURL-based EPUB Identifier (PEI) identifies an EPUB or EPUB 3 publication and allows EPUB/3 publications to reference other EPUB/3 publications. PEIs are valid HTTP URIs that can be distinguished from other URIs by their consistent prefix and created or used without network access. Additional optional metadata MAY be added to a PEI to identify a specific EPUB/3 publication, a single file inside the publication, or a single location inside a file. Finally, some PEIs MAY be resolved to an available resource using HTTP and redirects against a Persistent Uniform Resource Locator (PURL) service hosted by the IDPF. Notational ConventionsThe key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119]. Parts of a PEIhttp://purl.idpf.org/{epub-publication-identifier}?{key1=value1&key2=value2...}PEIs are made up of three components:
Creating a PEIGiven an EPUB or EPUB3 publication, a PEI is constructed by escaping the Publication Identifier from the EPUB/3 metadata and appending it to the address for the IDPF-hosted PURL service, http://purl.idpf.org/. The Publication Identifier, defined in the section "Publication Identifiers" of [EPUBPUB3], is the text content of the dc:identifier element referenced as the publication's unique-identifier. This text content MUST be stripped of leading and trailing whitespace before escaping following FIXME (this is often referred to as "percent encoding"). All PEIs MUST be valid Uniform Resource Identifiers (URIs) as defined in [RFC3986]. Additionally, PEIs SHOULD include a query component following a ? character that contains at least two sets of key-value pairs ("query parameters") from the EPUB/3 metadata:
All query parameters MUST be stripped of leading and trailing whitespace before escaping following FIXME. Future versions of this document MAY specify additional query parameters for identifying parts or versions of publications using different techniques. Identifying Part of a PublicationPEIs MAY include additional query parameters to identify a single file in an EPUB/3 publication and/or a single location inside that file. These query parameter keys are defined:
Identifying a Specific Version of a PublicationPEIs MAY include additional query parameters to more explicitly identify a precise version or instance of an EPUB/3 publication. These query parameter keys are defined:
Using PEIs in EPUBsValid PEIs MAY be used inside EPUB/3 publications to reference other EPUB/3 publications. For example, this XHTML markup could be used inside an EPUB/3 publication to reference another work: For more information on creating ebooks in the EPUB format, see <a href="http://purl.idpf.org/0132366991?title=EPUB+Straight+to+the+Point&creator=Liz+Castro">Liz Castro's EPUB Straigh to the Point</a>. A PEI-aware EPUB/3 Reading System MAY use this hyperlink to switch titles (if the user already had access to this title and an Exact Match was made) or to invoke a local or remote acquisition attempt or search (following deconstruction of the PEI, as described in the section "Reading PEIs"). Reading PEIsPEIs can be deconstructed and used by EPUB/3 Reading Systems or PEI clients by extracting and unescaping the Publication Identifier value. This value can then be matched against the Publication Identifier of available EPUB/3 publications. This is an Exact Match. If an Exact Match is not found, an EPUB/3 Reading System or PEI client MAY also choose to attempt a to locate similar publications based on available title, creator, and/or other query parameter values, if available. This is an Approximate Match. A user SHOULD be notified if an Approximate Match is returned. Some PEIs MAY have resources availabe via HTTP [RFC2616], as described in the section "PEIs and PURLs". Processing ExpectationsPEI creators PEIs MUST NOT require an active internet connection to create a PEI from an available EPUB/3 publication. PEI clients MUST NOT:
PEI clients MUST ignore unexpected query parameters in a PEI. PEI clients SHOULD preserve the entire query component, including unknown parameters, when storing or retransmitting PEIs. PEIs and PURLsThe IDPF will provide a PURL service for registering and maintaining HTTP redirects from valid PEIs. (FIXME some detail here.) Please see http://purl.idpf.org for more information. PURLs are discussed in more detail in [PURL-Overview]. Registering a PEI with the PURL serviceValid PEIs MAY have HTTP redirects registered with the PURL service. The resources returned by these redirects are out of the scope of this document. Existing PEIs registered with the PURL service SHOULD be maintained by the original registrant using the interfaces provided by the service. ExamplesA simple PEI identifying an EPUB publication with the Publication Identifier <dc:identifier id="pubid">urn:uuid:6C049021-7236-4866-82C6-6743B59B77A4</dc:identifier>: http://purl.idpf.org/urn%3Auuid%3A6C049021-7236-4866-82C6-6743B59B77A4 The same PEI with the recommended title and creator query parameters: http://purl.idpf.org/urn%3Auuid%3A6C049021-7236-4866-82C6-6743B59B77A4?title=PEI+Example&
creator=Creator1+Gda%C5%84sk&
creator=Creator2A PEI identifying a specific file in an EPUB3 publication: http://purl.idpf.org/0132366991?file=%2FOEBPS%2FePub-STTP-4.xhtml A PEI identifying a specific location inside a specific file in an EPUB3 publication: http://purl.idpf.org/0132366991?title=EPUB+Straight+to+the+Point&
creator=Liz+Castro&
file=%2FOEBPS%2FePub-STTP-4.xhtml&
fragment-id=toc-anchor-4&
text=The+mimetype+file+is+a+simple+text+fileA PEI with a number of optional parameters: http://purl.idpf.org/the_hound_of_the_baskervilles-AAH812&title=The+Hound+of+the+Baskervilles&
creator=Sir+Arthur+Conan+Doyle&
file=%2FOPS%2Fthe_hound_of_the_baskervilles-AAH812_chapter_03.html&
fragment-id=rw-p_39510-00001&
md5sum=24aeed0b82aadf061d88c7143dc6ca2bDOI and PEIDOIs can integrate seamlessly into PEIs. The registration, maintenance, and specifics of DOIs for EPUB/3 publication is out of scope. DOIs are described in more detail in [ISO-26324]. Valid DOIs MAY be used as a Publication Identifier, like: <dc:identifier id="pubid">urn:doi:10.1000/182</dc:identifier>`: This DOI would then be escaped as usual to create a PEI: http://purl.idpf.org/urn%3Adoi%3A10.1000%2F182 The difference would be that DOI-aware applications would recognize the urn:doi: prefix and retreive more information about the DOI after extracting the Publication Identifier from the PEI. TODO
ReferencesNormative References
Informative References
|
► Sign in to add a comment
While a PURL-based service has many advantages, there are many difficult (i.e. non-technical) issues that need to be settled before it could be included in the standard. Who controls the redirect urls? What do you do when ownership of rights is transferred? How do you manage regional rights issues? How do you ensure long-term persistence? What governance mechanisms will work. A thorough study of CrossRef? will give some hints as to the issues involved.
Putting bibliographic metadata on a link has its own set of issues. Here, a look at OpenURL, the most widely used bibliographic-metadata-on-url format is worth a look.
Both CrossRef? and OpenURL, separately and together, have succeeded as linking formats because they focused on the process of link creation. It seems to me that these proposals need to explicitly lay out a process by which inter-object links might be added to epub files in practice as well as how they might be resolved. Will links be created from text citations? If so, either an OpenURL-like or CrossRef?-like apparatus needs to be put in place. Will third parties be spidering a universe of file and inserting links? if so, a URI based Linked Data-ish system is more appropriate.
md5 hash on the whole file will not work in cases where the package is modified (e.g. when additional metadata is added to a file, when content is encrypted, when DRM license is written in rights.xml or when a signature is applied). It should be taken out and whole-file-checksum method should be specifically warned against in the spec.
For an interesting approach to internal linking, it's worth looking at the Times' Emphasis: Times Open
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml" > <head>
</head> <body> <h2>Requirements</h2> </body> </html> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml" > <head> </head> <body> <h2>Requirements</h2> </body> </html>