Phil Agre on the “backpack metaphor” in user interface design, from a 2001 issue of Red Rock Eater:
(5) Interfaces in this world might be organized according to various
metaphors, by analogy to the famous (if inconsistent and half-baked)
desktop metaphor in personal computing. One metaphor might be the
“backpack” in which individuals carry around the virtual documents
and devices that they use in various environments. We can easily
imagine someone sitting down in a cafe and spreading their paperwork
on the table, whether for their own use or to share with someone
else, such as a customer. The objects in the backpack would interact
with whatever devices they find in the vicinity, converting formats
where necessary or providing the user with options about which nearby
devices should be recruited for which purposes. The backpack metaphor
suggests an obvious set of operations: adding things to the backpack,
cleaning it up, securing it against theft, having it searched by the
security people, and so on, some of which will make sense for a given
application and some of which will not. (On the foundations of the
backpack metaphor, see the “Placeless Documents” project at Xerox.
On documents generally see David M. Levy, Scrolling Forward: Making
Sense of Documents in the Digital Age, Arcade, 2001.)
More on Placeless Documents from PARC’s Paul Dourish:
Placeless Documents are documents that are organized and managed according to their properties, rather than according to their location. Document properties can be things you already know about your documents, like that they’re published, or notes, or about the budget,or drafts, or source code, or important, or shared with your colleagues, or from your manager, or big, or from the Web, or… whatever suits you. Document properties can also be things that you want to be true about your documents, like that they are backed up, or replicated on your laptop, or can be purchased for a small fee. These latter properties carry the code to implement or interface with the desired functionality. Document properties are statements about your documents that make sense to you, and affect what you’re going to do with the documents.
The particular implementation of placeless documents, a system called Presto, was filesystem and not browser based; but a lot of the ideas look like they lead over neatly into what tagging systems like delcious do. In particular there’s the strong notion that the document properties you assign (“to read”, “due next thursday”) vs those that someone else assigns (“interesting”, “from PARC”) don’t have to match up and don’t have to interfere with each other. Presto went one step further and allowed some document properties to be inferred from document contents, as though delicious fetched each page that you bookmarked and analyzed it for authorship, date, or other bits of metadata. (That’s somewhat easier to do in a local filesystem than over the web, though systems like Google and Citeseer both cache local copies of documents to aid in retrieval)
Another good reference: Extending Document Management Systems with User-Specific Active Properties (Dourish99Extending).