Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4180 Silvia.Pfeiffer@csiro.au http://www.ict.csiro.au/

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4222 Conrad.Parker@csiro.au http://www.ict.csiro.au/

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4222 Andre.Pang@csiro.au http://www.ict.csiro.au/

This specification defines the Continuous Media Markup Language (CMML), version 2.0, an XML-based markup language for time-continuous data. It is a sister document to the specification of the Annodex annotation, indexing and hyperlinking format for time-continuous data. A CMML file is essentially a textual representation of an Annodex file. The tags of a CMML file provide for the creation of structured and unstructured annotations as well as hyperlinks and addressable named anchor points for clips of time-continuous data. Through its import tag, the CMML is also an authoring language for Annodex streams. The tag names in use in CMML are similar to the ones in XHTML. The 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 RFC 2119.

The Continuous Media Markup Language (CMML) specifies XML based markup for time-continuous data to allow it to become an integral part of the World Wide Web analogously to how HTML allowed text documents to become part of the Web. Therefore, format of the CMML derives much from XHTML. CMML allows to attach free-text annotations, metadata, captions and other textual information to clips of time-continuous data, thus enabling a timed textual representation of the data, which can be indexed by Web search engines. CMML also allows to attach a hyperlink to clips of time-continuous data, enabling Web search engines to crawl the content. This also enables users to surf seamlessly between time-continuous data and other Web resources, integrating clips of media into the browsing history of a Web browser. CMML also allows to attach a representative image to clips of time-continuous data, providing for a visual representation of the clip in conjunction with the textual representation as, for example, in the presentation of search results or in a table of clips. CMML provides for a "head" element to store information that concerns the complete time-continous resource, and a set of "clip" elements that each store information for a temporal subpart of the resource. The practical use of a CMML file is in conjunction with the Annodex exchange format. CMML markup can be interleaved inside an Annodex file or stream to allow a synchronised delivery of marked-up time-continuous data in a single stream between a Web server and a user agent. CMML has also been designed as an authoring language for Annodex bitstreams. It allows to describe the time-continuous data bitstream(s) that need to be multiplexed together to create an Annodex bitstream. This information is stored in the "stream" element of a CMML document. Such a document can be used to control the multiplexing process that creates an Annodex file. The following picture illustrates the multiplexing activity schematically; in reality, the stream tag is not preserved in its original form and some attributes are also made irrelevant during multiplexing. Details of how CMML markup is encoded in an Annodex bitstream are given later in this document.

-<------- | Multiplexing | v --------------------------------------------------------------------- |stream|head|clip_1| data packets |clip_2| data packets ... --------------------------------------------------------------------- ]]>

The CMML is technically fully specified through its DTD as given in the Appendix. The semantic meaning of each of the tags, their content and their attributes is specified in the following sections. The Appendix also contains an example of a CMML (instance) document. The file extension of CMML files is ".cmml". This document also applies for registration of the mime-type "text/cmml" for CMML files with IANA. In the meantime, "text/x-cmml" will be used. Please note that this document assumes that the reader has a fluent working knowledge of Extensible Markup Language (XML), Hypertext Markup Language (HTML), XHTML and the World Wide Web. Basic knowledge about the Annodex format is also assumed.

At the beginning of the CMML DTD, several parameter entities are defined that are used throughout the DTD as data types. This section gives a brief overview of them and refers to the relevant standards in which they are defined.

A "ContentType" specifies the media type and subtype of a document as defined in RFC 2045. It is used to specify the type of content that one input time-continuous bitstream contains. Examples are "application/annodex", "audio/x-speex", "video/x-theora", or "video/mpeg".

A "URI" is a character string that conforms to the specification of the Uniform Resource Identifier as defined in RFC 3986. A URI generally points to a Web resource. The URI time interval specification is supported for CMML and Annodex files. Also, direct addressing of clips as specified in the MIME type application part of this document is supported for CMML and Anndex files.

The "LanguageCode" defines a collection of constant strings that each identify a specific language as defined in RFC 1766. Examples are: en-au, de, x-klingon. Language codes are used to provide internationalisation support.

To provide international language support, the i18n entity draws together a language given by a "LanguageCode" in "lang" with the directionality of that language in "dir" given either as ltr (left-to-right) or rtl (right-to-left). "ltr" is the default.

There are three different time specifications in use in CMML: "Timestamp", "Playbacktime" and "UTCtime". A "Timestamp" is generally a name-value pair which defines a time point. The time point value is interpreted according to the time scheme given in the name. If the name is ommitted, it defaults to "npt:". Valid time schemes are the ones defined in the temporal URI specification. The "Playbacktime" entity is a data type that just specifies a SMPTE or a NPT time. It is therefore equal to the Timestamp entity without the UTC specification. The "UTCtime" entity is a data type that just specifies a UTC time without an identifier. UTC time is specified as in the Timestamp entity, but without the "clock:" identifier.

A CMML file is an XML instance document of the CMML DTD. An example is given in the Appendix. It starts with the usual xml directive and the DTD specification (see http://www.w3.org/TR/REC-xml#sec-prolog-dtd). The following is an example preamble:

]]>

The attribute "standalone" is set to "yes" when the only DTD that is used for the instance document is cmml.dtd. The encoding format specifies the character encoding that is used for the values of the attributes and elements of the CMML file. E.g. for languages that are non-latin based, such as most Arab and Asian languages, a simple character encoding like US-ASCII does not cover all the characters. The default "UTF-8" charset can accommodate for any and all. After the preamble, the CMML tag follows. A CMML file has a "cmml" tag as the root element. It embraces all the other tags.

]]>

The "cmml" tag encloses at most one "stream" element, exactly one "head" element, and as many "clip" elements as the document author requires. A "clip" element describes a section of the related Annodex bitstream. Attributes of the "cmml" element are the usual xml root tag attributes: the internationalisation attributes "lang" and "dir", an identifier "id" and a fixed namespace "xmlns". The internationalisation attributes specify the default language (language and directionality) of the complete CMML document. If not given, the language default adheres to the same rules as HTML, where the setting of the HTTP "Content-Language" header may specify the default language of a HTML document received over HTTP, or ultimately the user agent defaults and user preferences set the language. (see http://www.w3.org/TR/REC-html40/struct/dirlang.html) Every element has an "id" attribute. The value of the "id" attribute MUST be unique within the document. It allows to uniquely identify an instance of an element and address it.

The "stream" element contains information that is used for authoring Annodex bitstreams from existing time-continuous data. The Annodex bistream is created by multiplexing the bitstreams given in the "src" attributes of the "import" tags of the "stream" element together with the CMML annotations in a time-synchronous manner. The "stream" element describes the input time-continuous bitstreams that are to be multiplexed together on authoring the Annodex bitstreams inthe "import" tags. Its attributes describe other features of the Annodex bitstream such as the time mappings for the start of the file.

]]>

The "stream" element has no text attributes and thus internationalisation attributes are not required. The "id" attribute follows the default language specified in the "cmml" element. The "timebase" attribute contains a playback time in seconds associated with the first data packet of the Annodex bitstream. All other times in the CMML file MUST be calculated relative to this timebase. For example, a timebase of 300 seconds npt for a video file implies that the first frame is related to a play time of 300 seconds, and a clip with a start time of 350 seconds is to be included 50 seconds into the Annodex bitstream. If no timebase (or no stream tag) is given, the timebase defaults to 0 npt. The timebase can be given as a SMPTE or NPT time, or as a rational number as in 5/1300, but not as a utc time. The "utc" attribute associates a calendar date and a wall-clock time with the timebase. It therefore provides a mapping of the timebase to a real-world clock time and is given as a UTC time. If it is omitted, the start attribute in the import tag, and the start and end attributes in clip tags MUST NOT be specified as UTC times. The content model of the "stream" tag then proposes an arbitrary number of input bitstreams. These are described one by one in the "import" element.

A "import" tag contains information on one of the input bitstreams for the multiplexing process. It may also contain additional parameters to set up the Annodex encoder for each import bitstream.

]]>

The relevant bitstream (fragment) is referenced through the "src" attribute. The src is a URI and may thus also contain a time interval specification in URIs which narrows down the input file to that given subpart. That resource is multiplexed into the Annodex bitstream starting at the time given in the "start" attribute and ending at the latest at the time given in the "end" attribute. The "start" and "end" attributes are interpreted relative to the timeline of the Annodex bitstream. The internationalisation attributes provide the language of the import element's and the contained param tags' attribute values, such as the "id" attributes and the "title" attribute. The optional "title" attribute provides a chance to jot down a human readable comment on the source bitstream. This may e.g. be used in authoring applications for a more human readable display than the "id" tag which is really a key for identifying elements uniquely. The "granulerate" attribute contains the base temporal resolution in Hz of the input bitstream referred in the "src" attribute. It depends on the encoding format of the input bitstream and typically contains the framerate for video (e.g. 25 frames/sec) and the samplerate for audio (e.g. 44100 samples/sec), but may contain any rational number given with an integer denominator larger than 1 sec (e.g. 25 frames on 2 seconds). Each bitstream has its own granulerate dependent on its specific encoding. This attribute is implied as it can be determined automatically during the multiplexing process from the headers of the encoded media bitstream. For bitstreams without header, such as uncompressed audio, the author of the CMML file can provide the granulerate to the multiplexer in this attribute. The "contenttype" attribute specifies the media type of the input bitstream referred in the "src" attribute. It is optional as the media type can often be derived from the file name or file header of the media source during multiplexing. The "src" attribute specifies a URI to the input bitstream. Commonly used URI schemes are "file" and "http". For specifying temporal subsets of the input bitstream, use the time interval specification for URIs. The "start" attribute specifies a time in the output Annodex bitstream at which the media bitstream will be inserted. This time is specified with respect to the "timebase" attribute given in the "stream" element. The "end" attribute specifies a time in the output Annodex bitstream at which the media bitstream will stop at the latest. This time is also specified with respect to the "timebase" attribute given in the "stream" element. This attribute is not required when the full bitstream is used. The content model of the "import" tag then allows an arbitrary number of "param" tags to add as many descriptive parameter values to the mulitplexing activity as necessary.

A "param" tag is empty, but its attributes contain a name-value pair for describing the input bitstream in the parent "import" element. It inherits its internationalisation from that element, too, to avoid overhead. The "param" element is declared as follows:

]]>

The "name" attribute identifies a property name. It does not list legal values for this attribute. The "value" attribute specifies a property's value. It does not list legal values for this attribute. An example parametrisation is the provision of machine-processable low level meta information about the import bitstream such as a video's image height and width and framerate.

The CMML "head" element contains annotation information on the complete Annodex bitstream, for whose creation the CMML file is used. It therefore contains header-type information such as a title, style information, related documents and meta information describing the bitstream. The "head" element is declared as the following:

]]>

The "head" tag must contain a "title" tag. It may contain one "base" tag before or after the "title" tag and any number of "meta" tags at any position. The "%i18n;" attribute specifies the base language of the "head" tag's attribute values. The value of the "profile" attribute is a space-separated list of base URIs specifying locations of "meta" tag schemes such as the Dublin Core (see http://dublincore.org/). These schemes may be used in the "meta" elements of the "head" or the "clip" tags.

The "title" tag gives a descriptive title for the complete Annodex bitstream. It is not considered to be part of the presentation and should be displayed, e.g. as the title of the window that the Annodex bitstream is being displayed in. Exactly one title is required per document. The "title" element is declared as the following:

]]>

The "%i18n;" attribute specifies the base language of the "title" text.

The "base" element defines the base URI of the Annodex bitstream. All relative URIs of the bitstream get interpreted relative to this base. The "base" element is empty, but its attributes contain the base URI. It is declared as follows:

]]>

The "href" attribute contains the base URI. If the "base" element is omitted, the base URI of the Annodex bitstream is derived from the address through which the Annodex bitstream is accessed.

The "meta" element in the "head" element defines structured annotations for the complete Annodex bitstream. A "meta" element is empty, but its attributes contain the name-value pairs of a structured annotation. The "meta" element is declared as follows:

]]>

The "%i18n;" attribute specifies the language of the meta attribute and content texts. The "name" attribute identifies a property name. It does not list legal values for this attribute. The "content" attribute specifies a property's value. It does not list legal values for this attribute. The "scheme" attribute names a scheme to be used to interprete the property's value. The scheme can be located via the "profile" attribute in the "head" element.

A CMML file typically contains a number of sections given through "clip" tags. The CMML "clip" tag contains information about a section of the Annodex bitstream. This is expressed in a number of elements and attributes annotating, indexing, and hyperlinking the section. The "start" and "end" attributes are used to give the insertion time for the clip into the Annodex bitstream.

]]>

Any number of "meta" elements may appear in a clip, and at most one "a" element, one "img" element, and one "desc" element. Though "meta", "a", "img", and "desc" tag are given in a specific order in the DTD, their order is actually random. A "clip" element defines a unique identifying name for the clip in its "id" attribute. This name can be used in URIs that point either to the CMML file or the Annodex bitstream created from it, and allows to point straight at the clip. This may either be done as a URI fragment or URI query specification. The "%i18n;" attribute specifies the base language for all the clip's attribute values and content elements. The "track" attribute specifies the track that this clip belongs to. An annotation track is a set of clips that belong together from a semantic point of view. Clips in the same track must not overlap temporally. A default track must be available always. This track is the one a client (such as a Web browser plugin) will display by default. Other annotation tracks may be created by the document author to describe a more specific content. An example use are different annotation tracks for each speaker in an audio recording of a meeting or tracks of different languages. The "start" and "end" attributes specify the time range during which the clip element is defined. This time range is specified with respect to the "timebase" and "utc" attributes given in the "stream" tag. If the "stream" tag does not contain a "utc" specification, "start" and "end" times are not allowed to be given in UTC time. "start" is a required attribute because a clip without a start time is useless. "end" is optional and only required where clips cannot continue on to the following clip.

The "meta" element is specified above in the "head" section. While a "meta" element in the "head" tag provides meta information for the complete Annodex bitstream, the "meta" elements in a "clip" tag only provide meta information for the clip.

The "a" element specifies a link to a related Web resource together with some information on that related resource. The "a" element definition is very closely related to the xhtml "a" element definition with a reduced number of attributes as they make sense for time-continuous data.

]]>

The internationalisation attributes specify the language of the anchor's attribute values and of the anchor text. The "class" attribute allows to override style sheet defaults for this anchor instance. The "href" attribute specifies the location of a Web resource given through a URI. It thus defines a link between the current clip and a resource which the author believes to be connected closely to this clip's content. This might be a html page or another Annodex bitstream clip or an image etc. An "a" element without a "href" attribute is illegal and MUST be flagged or ignored. The text contained in an "a" element (i.e. the anchor text) provides a short textual description of the link specified through the "href" attribute. It explains why the connection between the current clip and the destination URI is made. It may e.g. encourage the viewer to follow the link to "Get more information on blah".

The "img" element specifies a link to a representative image for the clip. This image should be quite small as it is the representative image (known as "keyframe") for the current clip. This image may be used to visually summarise the content of the clip when a link to it is displayed, e.g. by a search engine or in a table of clips. The "img" element definition is very closely related to the xhtml "img" element definition with a reduced number of attributes as they make sense for time-continuous data.

]]>

The internationalisation attributes specify the language of the image's attribute values. The "src" attribute specifies the location of an image on the Web given through a URI. The "alt" attribute specifies alternative text to be displayed instead of the image as required e.g. for accessibility.

The "desc" tag contains a human readable, textual description of the content of the clip. The "desc" element is declared as the following:

]]>

For extracting a short text from the "desc" element as needs to be displayed in a table of clips or as caption, the first few characters of the description will be taken. It is therefore recommended to place a short meaningful summary sentence at the beginning of the description when authoring annotations. The internationalisation attributes specify the language of the text in the description and the "id" attribute a unique identifier for the element.

CMML is an annotation language that is meant to mark up any time-continuous data and be interleaved in a time-synchronous fashion with other time-continuous bitstreams. Therefore, CMML must be able to be serialised into a time-continuous bitstream of data packets. This is described in this section. CMML is serialised by having some initial header pages that set up the CMML decoding environment, and contain header type information. The content of a CMML bitstream then consists of "clip" tags. The "stream" tag is not copied into the CMML bitstream as it controls the authoring of the Annodex bitstream. Its information can be used in the encapsulation format. All of the CMML bitstream information is text. As it gets encoded into a binary bitstream, an encoding format has to be specified. To simplify things, UTF-8 is defined as the mandatory encoding format for all data in a CMML binary bitstream. Also, the encoding process MUST ensure that newline characters are represented as LF (or "\n" in C) only and replace any new line representations that come as CR LF combinations (or "\r\n" in C) with LF only.

The first header packet of a CMML logical bitstream is the CMML ident header. It contains all information required to identify the CMML bitstream and to set up a CMML decoder. It has the following format:

The fields in an CMML ident header packet have the following meaning: Identifier: a 8 Byte field that identifies this file to be of a CMML logical input bitstream. It contains the magic numbers: 0x43 'C' 0x4d 'M' 0x4d 'M' 0x4c 'L' 0x00 '\0' 0x00 '\0' 0x00 '\0' 0x00 '\0' Version major: 2 Byte short integer number signifying the major version number of the CMML format bitstream. Version minor: 2 Byte short integer number signifying the minor version number of the CMML format bitstream. When encapsulating a CMML bitstream, more fields may be added to this header as required by the encapsulation or exchange format.

The CMML secondary headers are a sequence of two packets that contain the CMML and XML "setup" information: one packet with the CMML xml preamble and "cmml" tag. one packet with the CMML "head" tag. These packets contain textual, not binary information. The CMML preamble tags are all single-line tags, such as the xml processing instruction (]]>) and the document type declaration (]]>). The only CMML tag that is not already serialized from a CMML file is the "cmml" tag, as it encloses all the other content tags. To serialise it, the "cmml" start tag is transformed into a processing instruction, retaining all its attributes (]]>), and the "cmml" end tag is deleted. The first CMML secondary header packet has the following format:

The second CMML secondary header packet contains the CMML head element with all its attributes and other containing elements and has the following format.

| +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ]]>

The data packets of the CMML bitstream contain the CMML clip elements. Their "start" and "end" attributes however only exist for authoring purposes and are not copied into the bitstream, but are rather represented through the time mapping of the encapsulation format that interleaves CMML data with data from other time-continuous bitstreams. This avoids contradictory doubly represented timing information. Generally the time mapping is done through some timestamp representation and through the position in the stream. A "clip" tag is encoded with all tags (except for the "start" and "end" attributes) as a string printed into a clip packet. The "clip" tag's "start" attribute tells the encapsulator at what time to insert the clip packet into the bitstream. If an "end" attribute is present, it leads to the creation of another clip packet, unless another clip packet starts on the same track beforehand. This clip packet contains an empty "clip" tag, i.e. a "clip" tag without "meta", "a", "img" or "desc" elements and no attribute values except for a copy of the "track" attribute from the original "clip" tag.

| +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ]]>

When mapping a CMML logical bitstream into Ogg, the serialisation as described in the previous section is used as a logical bitstream. The ident packet is extended by a few fields that are necessary for handling the time stamping of the content packets (i.e. the clips) for Ogg. Here is its format:

Fields with more than one byte length are encoded LSB (least significant byte) first. The additional fields in a CMML ident header packet for Ogg have the following meaning: Granule rate numerator & denominater: 8 Byte integer number each. They represent the temporal resolution of the logical bitstream in Hz given as a rational number in the same way as the fishead timebase field above. Granuleshift: a 1 Byte integer number describing whether to partition the granule_position into two for the CMML logical bitstream, and how many of the lower bits to use for the partitioning. The upper bits then still signify a time-continuous granule position for a directly decodable and presentable data granule. The lower bits allow for specification of the granule position of a previous CMML data packet (i.e. "clip" element), which helps to identify how much backwards seeking is necessary to get to the last and still active "clip" element (of the given track). The granuleshift is therefore the log of the maximum possible clip spacing. The default granule rate for CMML is: 1/1000. The default granule shift used is 32, which halfs the granule position to allow for the backwards pointer. The media mapping for CMML into Ogg is as follows: The bos page MUST contain the extended CMML ident packet. The first secondary header packet of CMML contains the xml preamble as described above. The second secondary header packet contains the CMML "head" tag as described above. The content or data packets for CMML contain the CMML "clip" tags each encoded in their own packet and inserted at the accurate time. The eos page contains a packet with an empty clip tag. If CMML is encapsulated in Ogg without the skeleton bitstream, it potentially loses time information. The timebase will then be mapped always to 0 and utc time mappings cannot be represented. It also loses all the message header fields which contain machine-readable meta information about the physical bitstream.

As CMML contains authoring information for Annodex bitstreams, a CMML instance document contains more than just the annotation information necessary for the CMML logical bitstream. It also contains control information to create the control section of an Annodex bitstream, i.e. the skeleton bitstream with its secondary header packets describing each of the contained logical bitstreams. Note that we only describe the creation of Annodex Version 3.0 bitstreams here. The authoring information stems in particular from the "stream" tag plus some specific information from the "cmml" tag. Generally, the "stream" tag's attributes contribute to the skeleton fishead packet, the "import" tag's attributes to the skeleton fisbone packets of each logical bitstream, and the "cmml" tag's attributes to the fisbone of the CMML logical bitstream. While the "cmml" tag is represented in full as a processing instruction in the secondary header packets of the CMML logical bitstream (see above), this is not the case for the "stream" tag. Therefore, this section also contains a description of what tags of the "stream" tag are not used inside an Annodex bitstream.

The skeleton ident packet receives the "timebase" and the "utc" field information from the "stream" tag. "Basetime numerator & denominator": if the "timebase" attribute is given in a CMML instance document, it MUST be represented in the skeleton ident header in the fields "Basetime numerator" and "Basetime denominator". It is converted from a possible NPT or SMPTE representation to a rational number to be stored in these fishead fields. Note the name change from timebase to basetime with Annodex version 3.0, which further versions of CMML will also mirror. "Presentationtime numerator & denominator": to be filled by the muxer appropriately, e.g. reusing the timebase values. "UTC": if the "utc" attribute is given in a CMML instance document, it MUST be represented in the skeleton ident header in the "UTC" field.

A fisbone packet for a logical bitstream is created through the authoring information of an "import" tag in a CMML instance document's "stream" tag. One "import" tag contains information on one particular logical bitstream in the interleaved bitstream and thus creates one particular skeleton fisbone packet. "Granulerate numerator & denominator": if the "granulerate" attribute is present in the "import" tag, it MUST be represented in the fisbone header for the respective media bitstream in the fields "Granulerate numerator" and "Granulerate denominator". The encoder MUST however ascertain that the values are sensible, and if it knows the accurate granule rate for a logical bitstream overrun the user input with the one that was used during creation of the interleaved bitstream. "Content-type" message header field: this attribute MUST be represented in the respective skeleton fisbone packet as a message header field with name "Content-type", as it signifies the MIME type of the media bitstream, providing for a decoding hint. If the user does not specify the "contenttype" attribute, the encoder MUST provide it during the interleaving process. "ID" message header field: if an "id" attribute is specified for an "import" tag, it SHOULD be represented in the skeleton fisbone header for the respecitve media bitstream as a message header field with name "ID", as it signifies a short identifying machine-readable string for the import media bitstream. User specified message header fields: if "name" and "value" attributes are specified in the "param" tags of the "import" tag, these SHOULD be represented in the skeleton fisbone packet of the respective media bitstream as a message header field with the given name-value pair. These fields are highly dependent on the type of media bitstream handled and it therefore depends on the encoding tool to make a selection of the parameters acquired. For example, an audio bitstream that contains speech in a specific language may be identified during CMML authoring through a param element with "Content-Language" name, and acquired into the media bitstream message header field of the same name.

A CMML instance document that specifies annotations in "head" and "clip" elements does not get to use the "stream" tag to provide encoding hints for its CMML logical bitstream. Its encoding hints come from the "cmml" tag and the "encoding" attribute of the xml processing directive. "Number of header packets": this field has a fixed size of 3 for the CMML specification given in this document. It counts the CMML ident packet, the XML preamble packet and the head tag packet. "Granulerate numerator & denominator": is fixed to the default value of "1/1000". "Content-type" message header field: the content type for the fisbone packet that describes the CMML logical bitstream is fixed at "text/x-cmml". "charset": if the xml processing directive contains an "encoding" attribute, this MUST be represented in the CMML fisbone packet as an addendum to the message header field "Content-type" as a charset. For example: "Content-type: text/x-cmml; charset=UTF-8". "ID" message header field: if an "id" attribute is specified for the "cmml" tag, it SHOULD be represented in the skeleton fisbone header for CMML as a message header field with name "ID", as it signifies a short identifying machine-readable string for the import media bitstream. "Content-Language" and "Content-Dir" message header fields: if the "lang" and "dir" attributes are given in a "cmml" tag, they MUST be represented in the fishbone packet of the CMML bitstream as message header fields with name "Content-Language" and "Content-Dir".

Here is a list of the attribute values of the "stream" tag and how they are being used: id: not used, as this attribute is only used to enable addressing of the stream tag the XML way (e.g. XPath). It is not relevant for the encoded bitstream and will therefore be lost on encoding. timebase: this attribute maps to the skeleton ident header fields "Timebase numerator" and "Timebase denominator". utc: this attribute maps to the skeleton ident header field "UTC". Here is a list of the attribute values of the "import" tag and how they are being used: id: this attribute may be represented as a message header field in the respective skeleton fisbone packet. lang, dir: not used, as these attributes signify the language and directionality of the human readable texts in the stream tag which are not acquired into the Annodex bitstream. granulerate: this attribute is used in the skeleton fisbone header fields "Granule rate numerator" and "Granule rate denominator" as well as for the "Presentationtime numerator" and "Presentationtime denominator". contenttype: this attribute is represented in the respective skeleton fisbone packet as a message header field with name "Content-type". src: not used, as this attribute only points to the location of the import media bitstream and is thus pure authoring information. start, end: not used, as this attribute only specifies the segment of the import media bitstream that is to be used during authoring. title: not used, as this attribute provides a human readable comment on the import bitstream for authoring purposes. Here is a list of the attribute values of the "param" tag list and how they are being used: id: not used, as this attribute is only used to enable addressing of the param tag the XML way (e.g. XPath). It is not relevant for the encoded bitstream and will therefore be lost on encoding. name, value: these attributes may be represented in the skeleton fisbone packet of the respective media bitstream as a message header field with the given name-value pair.

The decoding of an Annodex bitstream to CMML is roughly inverse to the encoding of an Annodex bitstream from a CMML file. There are some special cases to take care of, therefore the decoding steps are outlined here.

The data encoded in the CMML logical bitstream conists of the xml preamble, the "cmml" tag, the "head" tag, and the "clip" tags. These are fairly straightforward to extract. xml preamble and "cmml" tag: The xml preamble is constructed from the second header packet of the CMML logical bitstream. It contains the full xml preamble. It also contains the "cmml" processing instruction, which MUST be transformed back to a normal element and an end "cmml" tag be added at the end of the created CMML document. "head" tag: The "head" tag is constructed from the third header packet of the CMML logical bitstream, which contains the complete content of the "head" element. "clip" tags: The "clip" tags are constructed from the content of the CMML logical bitstream. Each packet contains a "clip" tag with all of the information except for the timing information. A decoder MUST take care to add the start time of each "clip" element into the "start" attribute of the respective CMML "clip" tag. The start time will be calculated from the granulerate in the CMML fisbone packet and the granulepos given in the respective "clip" Ogg packet. Empty "clip" tags should also be converted to end time attributes of the previous "clip" tag on the same track.

The creation of a "stream" tag is not necessary to extract the content of the CMML logical bitstream and thus a textual representation of the interleaved bitstream. However, if the Annodex bitstream has a non-zero "timebase" or a non-null "utc" time in the skeleton ident header, a "stream" tag will allow accurate time information in the CMML file and SHOULD be created with these attribute values. If a "stream" tag is created with the "timebase" and "utc" attributes, it is empty by default. A ripping application MAY however extract all the data bitstreams out of the Annodex bitstream into files, and then reference these files in the "src" attribute of "import" tags. Other attributes of the "import" tags MAY also be filled from the logical bitstreams: the "contenttype" attribute from the "Content-type" Message header field of the respective skeleton secondary header packet, the "granulerate" attribute from the Granulerate fields of the respective skeleton secondary header packet, the "id" attribute from a Message header field called "ID" if available, and "param" elements from all the remaining Message header fields of the respective skeleton secondary header packet, where the field name is stored in the "name" attribute and the value in the "value" attribute. A stream tag will thus roughly be created like this:

]]>

This section contains the registration information for the 'text/cmml' media type. While this media type is not approved by the IANA, 'text/x-cmml' may be used to identify CMML instance documents. To: ietf-types@iana.org Subject: Registration of MIME media type 'text/cmml' MIME media type name: text MIME subtype name: cmml Required parameters: none Optional parameters: charset (as in the text/xml media type). Encoding Considerations: as appropriate for the charset and the transport mechanism (see text/xml media type). Security considerations: see next section. Interoperability considerations: CMML is a free specification that is independent of any media encoding format. It is designed to provide interoperability with existing XML tools and systems. Its specification is not patented and can be implemented by third parties without patent considerations. Additional information: Magic numbers: none. However, CMML files start with the XML preamble as any XML document and will also have the string near the beginning of the file. File extension: .cmml Macintosh File Type Code: "TEXT" Intended usage: COMMON

There are two ways of hyperlinking via URIs into CMML files: via specification of a temporal interval or via specification of a clip. Both of these ways of addressing are supported for URI queries and URI fragments on CMML files. A server that is capable of supporting these URI queries indicates this by adding the X-Accept-TimeURI header field to its response header fields as defined in the temporal URI query specification. Specifications of a clip in a URI query via the clip's name are regarded as an alias for a time offset. Therefore, a server that supports CMML temporal URI addressing MUST also support the named addressing.

For the purposes of URI queries on CMML files, it is assumed that the query string takes the format of a CGI query string. The Common Gateway Interface, or CGI, is a convention for external gateway programs to interface with information servers such as HTTP servers (see http://hoohoo.ncsa.uiuc.edu/cgi/). This query string is expected to be interpreted by the HTTP server to return a valid CMML file that differs from the original CMML file only by reducing the set of clip tags to the specified interval. Temporal query parameter specification: Addressing of temporal intervals of CMML files is possible through specification of temporal query intervals in URIs. An example is the following URI: http://example.com/sample.cmml?t="npt:4" , which in the case of CMML relates to the last clip whose start time is just before the given temporal offset and all the clips thereafter. Clip specification: Addressing of a clip is possible through specification of the clip's id attribute value. The BNF for such an id URI query parameter is: All id-name specifications map onto a start and end time. Specifying just an id-name maps to just the start and end times of that clip. Specifying an id-name with a "/" maps to the document starting from the beginning of that clip until the end of the file or stream. Specifying an id range is inclusive and maps to the start time of the first clip and end time of the second clip. Overlapping time intervals MUST be interpreted by merging the intervals into one. It is not valid to give several temporal URI query parameters in one URI query. They all need to be wrapped into a single specification. Examples for URIs containing id queries are: http://example.com/sample.cmml?id="dolphin" --- sample.cmml is transferred from the start time of the "dolphin" clip to the end of the file/stream. http://example.com/sample.cmml?id="dolphin/" --- sample.cmml is transferred from the start time of the "dolphin" clip to the end of the file or stream. http://example.com/sample.cmml?id="dolphin/goldfish" --- sample.cmml is transferred from the start time of the "dolphin" clip to the end of the "goldfish" clip. Note that id attribute values of all elements have to be unique throughout a XML file (and thus also throughout a CMML file).

For the purposes of URI fragment specifications on CMML files, it is assumed that the fragment gets interpreted by the HTTP client after the retrieval action. The HTTP client is expected to restrict the usage of the resource to the specified interval. Temporal fragment specification: Addressing of temporal intervals of CMML files is possible through specification of temporal fragments in URIs. An example is the following URI: http://example.com/sample.cmml#t=npt:4 . This then relates to the last clip whose start time is just before the given temporal offset and all the clips thereafter. This may e.g. be useful to do a zoom into a retrieved CMML resource. Clip specification: The values of the id attribute of the clip tags can be used for addressing media clips directly through fragment identifiers as in http://example.com/sample.cmml#dolphin.

As CMML is a markup language created by using XML, the same security considerations that apply to XML, apply to CMML. As the CMML is an authoring language for Annodex bitstreams, there is no executable code attached to this language. The implementation of a multiplexer to actually create an Annodex bitstream must be careful when handling input bitstreams, which are binary data.

draft-pfeiffer-cmml-01: CMML version 2.0: changes to the tags to make CMML more similar to XHTML, in particular from "media" to "import", the introduction of an "img" tag, and the the change from an "a" tag that covers all the other elements to a "clip" tag, reducing the "a" tag back to its HTML meaning. draft-pfeiffer-cmml-02: CMML was not changed - only the media mapping into Annodex was adapted because the binary format had changed substantially.

Harvard University

29 Oxford Street Cambridge MA 02138 US +1 617 495 3864 sob@harvard.edu

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 timbl@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 timbl@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 timbl@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 howcome@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 bbos@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 bbos@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 howcome@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 chris@w3.org http://www.w3c.org

World Wide Web Consortium

MIT Laboratory for Computer Science 545 Technology Square Cambridge MA 02139 US + 1 617 253 2613 + 1 617 258 5999 ij@w3.org http://www.w3c.org

World Wide Web Consortium

Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge MA 02139 US +1 617 253 5702 +1 617 258 5999 timbl@w3.org

Day Software

5251 California Ave., Suite 110 Irvine CA 92617 US +1 949 679 2960 +1 949 679 2927 fielding@gbiv.com

Adobe Systems Incorporated

345 Park Ave San Jose CA 95110 US +1 408 536 3024 LMM@acm.org

Columbia University

Dept. of Computer Science 1214 Amsterdam Avenue New York NY 10027 US schulzrinne@cs.columbia.edu

Netscape Communications Corp.

501 E. Middlefield Road Mountain View CA 94043 US anup@netscape.com

RealNetworks

1111 Third Avenue Suite 2900 Seattle WA 98101 US robla@real.com

UNINETT

Pb. 6883 Elgeseter Trondheim 7002 Norway Harald.T.Alvestrand@uninett.no

Innosoft Internationl, Inc.

1050 East Garvey Avenue South West Covina CA 91790 USA ned@innosoft.com

First Virtual Holdings

25 Washington Avenue Morristown NJ 07960 USA nsb@nsb.fv.com

University of California, Irvine

Department of Information and Computer Science Irvine CA 92697-3425 USA ejw@ics.uci.edu

Fuji Xerox Information Systems

KSP 9A7, 2-1, Sakado 3-chome, Takatsu-ku Kawasaki-shi Kanagawa-ken 213 Japan murata@fxis.fujixerox.co.jp

The Society of Motion Picture and Television Engineers

595 W. Hartsdale Ave. White Plains NY 10607 USA smpte@smpte.org

International Organization for Standardization

1 rue de Varembre Case Postale 56 Geneva 20 1211 CH central@iso.org

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4180 Silvia.Pfeiffer@csiro.au http://www.ict.csiro.au/

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4222 Conrad.Parker@csiro.au http://www.ict.csiro.au/

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4222 Andre.Pang@csiro.au http://www.ict.csiro.au/

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4180 Silvia.Pfeiffer@csiro.au http://www.ict.csiro.au/

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4222 Conrad.Parker@csiro.au http://www.ict.csiro.au/

Commonwealth Scientific and Industrial Research Organisation CSIRO, Australia

PO Box 76 Epping NSW 1710 Australia +61 2 9372 4222 Andre.Pang@csiro.au http://www.ict.csiro.au/

]]>

Read more about fish This is the introduction to the film Joe made about fish. Here, Joe caught sight of a dolphin in the ocean. More video clips on goldfish. Joe has a fishtank at home with many colourful fish. The common goldfish is one of them and Joe's favourite. Here are some fabulous pictures he has taken of them. ]]>

XML tags and their content used to describe a document. The task of authoring mark-up for a document thus creating a Web resources. The task of linking from one Web resource to another. When a link contains a fragment offset into a resource, this is called "deep hyperlinking". A section of a time-continuous document covering some temporal interval. The task of identifying index points or clips for time-continuous documents. A set of clips representing semantically correlated annotations of a time-continuous resource. A specific file format for storing annotation, hyperlinking, and indexing information in annotation tracks and multiplexed together with the time-continuous documents they describe. A sequence of data containing samples of a time-continous document. A file containing time-sampled data in a temporally sequential manner.

Annotated and indexed bitstream format. Continuous Media Markup Language. Cascading Style Sheets. Document Type Declaration. eXtensible Markup Language. World Wide Web. Unified Resource Identifier.

The authors greatly acknowledge the contributions of Zentaro Kavanagh, Andrew Nesbit and Simon Lai in developing this specification.