Media resources, such as plain text files, images, audio, and video recordings, are published to the network at a permanent address as immutable source documents.
Documents, which may reside on- or off-line, are dynamic compositions of network media fragments; Media fragments are not stored inside of documents — they are referred-to by the document and loaded into the viewing context at display-time. This process is called transclusion, and is analagous to how audiovisual media are treated in HTML. However, the important difference between a xanalogical system and the Web is that all media, including text, is transclusible.
Links are logical structures that exist independently of documents. They are imported and overlaid onto viewing contexts, and may link any number of documents and media fragments into diverse topologies.
The "browser" is a multi-document interface with parallel viewing of texts and visible connections between linked media.
In order to facilitate a system of this nature, we believe that
All Web media must have a UNIFORM METADATA INTERFACE which provides not only technical data about the media, but authorial, editorial, curatorial, and other sociocultural information such as titles, descriptions, authorship, ownership, usage rights, and so-on;
That this metadata interface must also serve as a means of DISCOVERY for additional interfaces in order to foster a move away from Web documents as downloadable files toward a Web of extensible media objects;
And that one of these additional interfaces must facilitate DEEP ADDRESSING and FRAGMENTARY RETRIEVAL of media.
Furthermore, we believe that Web media should have the capability of recording its relationships to other media, and that it should be able to respond to queries about those relationships. To that end, we propose a system of automatic LINK REGISTRATION, and a corresponding LINK QUERY INTERFACE.?link