A trail is worth a thousand thoughts

In Vannevar Bush’s 1945 Atlantic Monthly article titled As We May Think

(see www.theatlantic.com/unbound/flashbks/computer/bushf.htm ) he suggested that instruments are at hand (such as the Memex machine described in the article) which, if properly developed, will give man access to – and command over – the inherited knowledge of the ages.

I would like to explore how close we are to this vision some 45 years later. First, I draw your attention to Section 7 where one of the most useful functions of the Memex is described – the ability to develop associative links between two previously unrelated documents.

He describes the construction of a “named trail” which I like to call “Vannevar Trails” or “VTs.” A VT is separate from the documents referenced. It consists of one or more items, with each item forming a link between a pair of documents via some unique code assigned to each document, and having some code word or word assigned to each item.

If a document is referenced, its VT also becomes visible, and the reader may reference the other linked document via the VT. All items in the VT can be reviewed, hence retracing the thinking of the original researcher or author who put this information together.

This scheme cannot be implemented well using HTML. A page developed by an author may be linked to any number of other pages, via the URL scheme. The pages referenced may represent the entire document (as in the Atlantic Monthly link above), or may represent some portion of the document. See www.isg.sfu.ca/~duchier/misc/vbush/vbush7.shtml ). This is a different copy of the same article that allows reference to section 7 directly. In this example, the original document was split into multiple HTML pages, each with a unique URL and hence could be referenced via the code Section        7.

Alternately, using HTML we could have added labels to the various sections of the original document and referenced the section via this label. For example if section 7 were modified with the label Section 7 , then Section 7 would take you directly to Section 7 of that document.

However, note that to accomplish this we have to modify the original document. Typically this is not achievable unless you are the author – and even then any document can be referenced in whole or in part by an arbitrarily large number of documents, rendering making a reference into a document via HTML impractical. As well, a page has no knowledge whatsoever of references being made to that page.

In SGML, a product called Panorama Pro from Softquad allows the construction of a separate file to reference any SGML document without modifying the original document. The reference, or collection of references, exists as a separate document encoded, according to the HyTime standard that provided explicit navigation into the structured document. For example, you might reference the 22nd to 43rd characters of the fourth paragraph, in the second section of Chapter 4. This particular application allows links between multiple annotation files and a source document without modifying the original document.

An interesting application called Thirdvoice (see www.thirdvoice.com) implements something similar. This Web-based application allows any user to annotate any Web page, start up dialogues around information on that page and make every word on the page “live” in the sense that if there is information about anything on the page, it can be made visible to the browser. Note that the referenced page viewed with Thirdvoice now appears to have knowledge about what pages are linked to it, all without modifying the original document. I have created a discussion group around the Atlantic Monthly article into which readers may contribute.

In XML, the Xlink and Xpointer standards (currently in Working Draft stage) offers a way to establish two way links between pages. Unlike the ThirdVoice solution, the Xlink solution would be non-proprietary. Xlink allows the construction of complex structured links between two or more “resources” associated with the same link item. These links may be in-line links (like the HTML lining mechanism) or links existing in an external file. Given these links may be structured and quite complex, we start to see technology that will allow the implementation of these “Vannevar Trails.”

W. Hugh Chatfield, ISP, is president of CyberSpace Industries 2000 Inc. in Ottawa. He is engaged in XML/SGML consulting and teaching and can be reached at hchatfie@attcanada.ca.