This article – Google’s War on Hierarchy, and the Death of Hierarchical Folders – talks about the issue of organizing filesystems for human interfaces, specifically it talks about the difference between Microsoft’s desktop and Google’s search paradigms.
It seems to me that the search paradigm, especially when used in this context to replace hierarchy, is anti-browse and pro-peruse to the extreme. When I know what I’m looking for, I can find it, but when I want to wander through files looking for something that grabs my attention something is missing. I say this having used Google’s and Microsoft’s desktop search tools, Beagle on Linux, and Apple’s Searchlight on a daily basis over the last year or so.
It also seems to me that it is better to have your files clustered by some sort of topic or metadata when using these kinds of systems, so that you get semantically applicable results when looking for search criteria that are vague or have multiple definitions or contexts. To do that you need metadata and clustering capabilities. And you need new metaphors for the information space.
I think a lot of work is happening in this area, but just not being well distributed. In my vision of a last generation GUI (some sort of ‘peak’ of non-Voice User Interface) I see a lot of things that are visually stunning and richly informative. Things like a zoomable environment, like Jef Raskin’s Humane Interface, or the city metaphor of tdfsb. I also really like the piles or stacks metaphor and the concept of smart folders in the filesystem and GUI, as well as the idea that those smart folders would be dynamically and persistently up to date with files on the local hard drive and on network accessible storage resources.
I’ll admit it, I like the visual nature of fictional GUIs and think that the kind of stuff shown off in Cyberpunk novels (immersive VI, ICE, et al.), Hackers, the Zion command in the Matrix films, the investigation desk in Minority Report, and the visual programming of Swordfish is cool, even if it isn’t probable with today’s technology. I think that GUI’s should be moving in those directions without allowing the CLI to wither and without foregoing a simple GUI for low-power systems such as Blackbox. I think that there isn’t simply a a competition between hierarchy and search, like this article seems to be implying, but rather one that is about finding the right balance between browsing and perusing.
Perusal is very important, and search is bettered by previous perusal. When the full text of the file system is indexed, finding every occurrence of a word is relatively easy. When the full file system is indexed with formal metadata like RDF, finding every occurrence of a predefined concept is relatively easy. When the full file system is indexed with informal metadata like folksonomic tags, finding every occurrence of a social concept is relatively easy. When the full file system is indexed to discern a particular pattern in media files – like images or music – then matching that pattern is relatively easy. Indexing, whether it is done by humans or by machines, is the prerequisite for search.Browsing, on the other hand, is a different sort of activity. Where perusal might be akin to buying, browsing is akin to shopping.
Browsing requires a visual metaphor that can be wandered through, like a hierarchical tree. Better metaphors exist, but the tree has persisted. I associate browsing with activities like wandering the stacks at a library or bookstore; it is an experiential, sensual activity. This implies looking to find something but not looking for something in particular. It implies the wandering of the mind, making connections between things, reviving memories, and creating new ones. This is, in some ways, a prerequisite for human indexing. Browsing is necessary to computing interfaces, and no amount of search technology can compensate for a wander through the data stacks.
So, anyhow, the ends are that we need better user interfaces, and we need better data management, and to get them we need new or better metaphors for the way we look at bits and bytes.