[conica]

Izvestaj je napisao Dennis Forbes

Uvodni tekst:

I recently had a need for a mid-sized amount of real-world data, which I required for testing purposes on low-end hardware (testing and demonstrating some of the new functionality of SQL Server 2005). I wanted something that wasn't confidential, which excluded the easy choice of using business data, and I refrain from using artificial data. Around the same time I happened across the requisition process for the .COM/.NET and .EDU TLD zones, so I made a request for access.

Soon enough I had the 3.5GB of .COM domain names, along with 650MB of .NET, loaded into the database (although for all results in this entry I only included the .COM TLD, for the data as of 2pm on March 28th, 2006. I'll analyze the other ones at a future date). It was a great foundation for a lot of tests and demonstrations, and served my original goal admirably. I didn't stop there, however; Curiousity led me to do some basic analysis to see what sorts of domain names are registered, and how saturated the registry really is.

Note that these are the Verisign distributed zone files, and do not include entries that have no nameservers configured, or which are in a hold state. While those comprise a very small minority of domain names, it does skew the results a bit. To improve accuracy when the sample set is small, for some of the tests I have validated the positives using the WHOIS infrastructure (for instance the domain file had several two letter sequences as being "available", and a dozen three letter sequences. All of them were the result of a hold state, or no nameservers configured). For aggregate results where it was inapplicable, I've filtered international domain names (IDN) from the results (prefaced with xn--).