the «Media Art Net» defines itself as mediation («Vermittlung») and purely verbal contextualization of items in the Internet, according to McLuhan's law it mirrors an old media in the new one (the content of a new medium is always the previous one): it conservatively sticks to the archival paradigm, which is based on inventories and linkage structures between files. Any archival linking is a freezing of relations, even if it opens up alternative links. Let us instead try to conceive digital data an-archivally.

Search engines as agents of mapping (the generative archive) confront us with the difference between browsing and searching. Crawling through the web using hyperlinks is different than targeted search operations. Maps without textual annotation are of no use to most users. When it comes to the text/image relation in searching-as-mapping, the technical option is to no longer exclusively subject images to the alphabet. Increasing use is now being made of graphical searching devices on maps (as opposed to the subject classification of «Media Art Net», which still follows the encyclopedic order). The ‹iconoclastic› option is a random audio-visual search for links (by similarity, by

digital association, connectivism, fuzzy logic). Which is the model: the ancient ars memoriae (the spatial order of images) or the archive (non-visual, logistical, data-based)? Early catalogues and inventories of works of art in the seventeenth century ordered works of art not according to intrinsic («history of art,» schools of painters), but purely external values: placed in real space (on the walls of a collection) according to formats—pure storage economy. Is the CD-ROM a contemporary form of such non-linear cataloging? [11] «Media art must be transmitted in a multi-media way» (Dieter Daniels and Rudolf Frieling in their editorial [LI]), since it is by definition timebased or process-oriented. It belongs to this specificity that digital audio-visual data can be addressed (= controlled) down to the smallest sample/bit/pixel; thus resulting in new search options with various access points.

Thus we arrive at literal bit mapping. «Image files contain basically a bit map; that is a long string of bytes […] each of which describes an individual pixel of the image. ‹Metadating the Image› is a subset of data knitting. We can tease out some indications of image structure and semantics (for instance, we can find all