I've decided to write about my bookmarking and history ideas here now.
The conceptual model is that of a carousel. It sounds extremely corny but could be good.
The idea is that a browser's history is stored in a list of each web page visited. This must include pages that were backtracked upon so that reverses in direction are not excluded. This means that a stack model of recalling visited sites will not work.
When each page loads, a thumbnail of the page is taken and stored in the browser's cache. When the user wants to examine their history, the page they are on zooms back slightly to show a long curving line of web pages (i.e., the history). It zooms back to ensure that the user is given visual cues as to where they are now.
The list of pages is not linear as this would be tricky to work out. My own studies indicate that most history searches are for fairly recent documents so these might be best if they are larger than earlier ones. They are also easier to click because they have a bigger surface area to click on (Fitt's law applies here). Obviously, when clicked on, they zoom up to take the entire page.
This is all well and good for selecting recent pages, but what about pages several hundred or thousand items back? Users can scroll along the list - as they go back in time, the newer pages zoom forward and past the users screen and out of sight while the older ones scroll nearer and become larger and easier to view - users can then make better judgements as to content. Scrolling forward in time does the reverse - users then have older pages receding into the distance while newer ones zoom back in to view.
The pages stand upright on a floor - this floor has indicators of date and possibly time. These help to provide a definite clue as to when pages were accessed which may help search. Pages that are bookmarked have a distinctive border that makes them stand out just in case the user wishes to select a bookmarked page. The user can perform a text query on the collection. All relevant documents then "jump up" a little higher than their sisters to allow the user to see which ones match their query along with another distinctive border (though different to bookmarked ones). They can easily select them, but also see them in relation to nearby history. This seems to tie in well with knowledge about navigation and how other pages may indicate proximity to the required target.
When a document is selected, the history will be searched for other times that the same document was accessed and these will also stand out from the others. This may be useful for documents that change and the user requires information only available from an old page.
How does this differ from other ideas?
My idea uses visual navigation. Web pages are also visual which helps matching. The temporal order in which documents were visited are stored. Because the design is not linear, more pages can be fitted onto the screen. There is also a bias in favour of the most recent documents. The exact date and time in which documents were visited is also visible which works better than "hazing" them - this gives a rough indication as to age, but not an exact idea. This differs from the traditional concept of a zooming interface which from my own work has problems of navigation (i.e., people getting quite lost!), confusion with 3d interfaces etc. This has a 3D effect but is actually a tuned 2D search - maybe more like Marr's 2.5D idea of the human visual system. Finally, I think the concept of moving backwards and forwards in time is appealing and easy to understand for most people. With a few jogs of memory of their browsing history, I think many would appreciate it.
Plus, it might also be fun to examine browsing history! Having spent a lot of time working in information search, it is quite surprising how much information one person can cover!
For really cool stuff, having eye tracking to navigate the history would be good. Look left at the older items to move back in time, and right to navigate forwards. Perhaps up to select the document underneath.