The database is finally reaching significant size so I decided to start playing with some of the things that I’ve been planning to do when a critical mass of spurls were reached.

The new addition is very small on the surface, but I suspect it may become a favorite feature for a lot of Spurl users: suggesting pages based on category contents. To see the related pages for one of your categories, open the category (in the main window, not just the Spurl bar) and click the “related spurls” link besides the category name at the top. Spurl will fetch 20 relevant pages, based on other users’ categorization.

The results are better than I even hoped for. If you choose a category with 5-10 or more links, the “related spurls” usually turns up a pretty impressive list of pages. The better defined the category is and the more pages in it, the better the results. I hope you will agree with me.

Addressing this from a more philosophical point, I’ve always had problems with the “one-size-fits-all” category structure of web directories such as Yahoo, dmoz, etc. Even with the cross referencing (the @-sign after the category names), it does not serve its purpose. I know only a handful of people that ever use these directories – people simply have so different mind-models of the world that categorization that makes perfect sense to one user, the next one finds totally irrelevant.

At the same time users tend to group related things together even though the exact hierarchy is totally different. So the thought was: What if I could simply browse my own category structure and yet find pages that I did not put there myself? Every Spurl user has put up a unique category structure – and believe me, there are as many such structures as there are users and people have fundamentally different approach in how they organize their bookmarks – yet there are often close matches between the contents of two categories even though they have different names, in different languages and are a part of a totally different category structure.

This way, everybody can use their own structure, yet tap into the power of the user base – and it will most likely scale nicely. The more users and categories, the closer matches this method will find.

All and any comments are welcomed, good or bad.