The World Wide Web is a pyramid of old conventions-- one-way links, exposed directories, the equating of one document with one base file, embedded thornyscript "markup,", and paper simulation with fixed point sizes (non-zooming WYSIWYG). To this the Andreessen browser added a salad of pseudo-random features, and XML now imposes hierarchy on that salad. Innumerable second-order conventions and tricks deriving from these accidental properties include weblogs, search engines, spy gifs, and of course "websites" themselves.

For shopwindows this may be fine; but I believe all these conventions are destructive of the true concerns of electronic literature: the transmission and presentation of ideas, stable publication, clean version management, and the ongoing work of authors, editors and scholars.

I will present a mix-and-match design for the handling of documents, content and interconnection which I believe is more appropriate for most human beings and most forms of text work. It is intended to embrace all documents into a large-scale, zoomable literary fabric-- with quotations connected to their origins, overlays creatable by anyone, and with embedded markup and imposed hierarchy converted to more neutral representations. Unlike its predecessor designs, DeepLit is interoperable with the Web's one-way pointers to pigeonholes and the Andreessen browser.