jmccormac - 6:55 pm on Apr 24, 2003 (gmt 0)

One of the biggest problems I have seen (especially with Dmoz-like directories) is that they grow organically rather than logically. As a result, the directory structure becomes more and more confusing for the user. This tends to drive users towards the search engines. The number of entries in the directory also has an effect on usability and navigation. From having to reverse engineer the Dmoz RDFs to build a simple MySQL/static webpage version, I can appreciate some of the logic of the Dmoz people but I don't think that it translates well for the users.

One of the main weakpoints of Dmoz is in the country sections. The number of listed sites is rarely more than 25% of the sites related to that country. According to Dmoz, it has 144795 UK websites (combined .uk/com/net/org/info etc). I am currently finishing pre-indexing UK com/net/org websites (the current count is running at 1237162 websites with under probably another 100K to go) and it seems that the number of excluded websites is really in the millions when you take the .uk websites into consideration. There is of course a qualitative argument here - what is worth including? However because Dmoz depends on user submissions, it is always going to be lagging behind the real world. The Dmoz dataset may be good for an initial boost but it alone does not provide a good enough data feed. A new directory would have to have other sources of new websites. The same 'lack of coverage' applies to all the countries that I have tracked. (My work involves tracking com/net/org website usage for the main Europe countries at the moment.) It may be possible to create a network of good country specific directories rather than one huge generic directory like Dmoz or Yahoo.

The business model of a brand new directory would be the important aspect to get right.

Regards...jmcc