October 28, 2003

Top Level Domain Follies

Bret Fausett quite reasonably argues that ICANN's TLD (Top Level Domain) "test bed" is dead.

I don't think that there ever was much life in that test bed.

I was trained in the hard sciences - mainly chemistry and physics.  And I spent some of my undergraduate years doing research on high input-power chemical lasers.  I also spent time in the soft sciences where I did research on patterns of urban mobility.  In all of this work we used a technique called "the scientific method" - it involves observation, formulation of hypothesis, predictions based on the hypothesis, and experiments to test those predictions (and indirectly the hypothesis.)

ICANN never really followed any process, much less one as structured those used in the hard sciences, to focus its observations of the behavior of new TLDs.  ICANN's information gathering was never better than ad hoc.  And there were neither hypotheses, predictions, nor experiments.  ICANN's TLD test bed process was not scientific; quite the opposite: it was chaotic and arbitrary.

Nor was ICANN's test bed process particularly useful for the creation of a body of data that might be useful for an unscientific after-the-fact inquiry.  ICANN's data gathering, even when it was performed, was mainly of business information that has no apparent relationship to the stability of the internet's domain name system.  No records were made of actual DNS activity and behavior as the new TLDs were being deployed.  No measures were made of the accessibility or usability of those new TLDs.

(It wasn't that there was not interest - Louis Touton and I wanted to quantitatively monitor the cross-fade of queries away from the old .org servers and onto the new ones as part of the Verisign-to-PIR transition of .org.  However, that effort was too low on ICANN's list of priorities and thus a golden opportunity to observe DNS behavior in the wild was lost.  Such data would have been invaluable when trying to comprehend the impact of a future planned or unplanned operational transition of a large DNS zone.)

To give but one example of something that we ought to have investigated:  There are thousands upon thousands of web page forms out there on the net that have Javascript or cgi-bin programs that do not accept top level domain names with four or more characters.  The people who operate the TLDs with such names have pointed out repeatedly that this limitation has substantially hindered the usability and acceptance of their offerings.  Yet, ICANN does not even have this problem listed as something that is to be investigated in the "testbed", and after several years, ICANN has not even a qualitative estimate of the extend of this problem, much less initiated a curative initiative to try to inform the web community that these limitations on web forms are overly restrictive.

ICANN's TLD "test bed" has little value except as a body of anecdotal data.

I agree with Bret that we should abandon the pretense that there is a "test" in progress or that there ever really was a "test".

It is high time for ICANN to move forward on new Top Level Domains - and not merely of the kind that, even in the absence of real tests, have shown little, if any, evidence of cognizable benefit to the community of internet users.

Posted by karl at October 28, 2003 1:17 AM