ldg, A couple threshold points in which you've misunderstood my analysis:
1) I don't take Yahoo listings to report that a domain is in use. That wouldn't be a good method -- it would be quite error-prone, as you say -- and it's not a method I used. Where specifically did you take the article to say I used such a method?
2) It's certainly true that many large companies hold large numbers of domains. But the fact that they do so, and presumably sometimes then release large numbers of domains, doesn't mean my research as to their primary domains is in any way skewed. What specifically makes you think the data "could be skewed by tens of thousands"? What is that even supposed to mean, in a sample of 1000 Fortune 1000 firms?
3) I didn't use Yahoo for search purposes. Instead, I used the Yahoo Directory for a sense of well-known sites on the web, and sites becoming sufficiently well known to be added to Yahoo. Certainly the task of devising a hierarchical taxonomy of the web is a difficult one, and I'm sure Yahoo would be the first to admit they could do that much better of a job with more resources, etc. But I don't think some errors in the Yahoo directory make it useless for the purpose for which I used it. Quite the contrary, it remains the case that the web's best-known web sites are listed in Yahoo, and that less well-known sites (even when listed in Yahoo) tend to be omitted.
4) All the discussion above, and the brunt of your comments, seem to discuss Alternative Perspectives on Registrar Market Share. Concerns about the response pool to the registrar survey would seem to be off-point as to the market share article. Certainly it's been none too easy getting responses from many registrars -- a failure for which I ultimately bear responsibility. But that's neither here nor there as to the other article.
Now, as to the substance of your comments:
You're right that my analysis puts great emphasis on whether domains are in use for providing web content. And you're right also that there are plenty of domains not used in this way. The key question, though, is whether my analysis is affected by the fact that some domains are used for purposes other than the web. I believe it is not.
I've tracked certain statistics (e.g. rate of usage for providing web content) according to registrar. It turns out, my statistics report, that domains registered through some registrars are considerably more likely to be listed in Yahoo than domains registered through other registrars. The reasons for this result are pretty clear -- different pricing policies, different UIs for bulk registrations, perhaps even different tolerances for allegedly-illegal or -infringing registrations. I agree that, lurking in the background, there are domains used for puroses other than the web. But is there reason to suspect domains with such usage are correlated with registration through certain registrars? If so, that's a problem for my results. If not, no problem.
My casual sense, based on intuition and borne out in casual investigation, is that there's no big difference across registrars as to domains used only for email or other non-web applications. My sense is that email-only affects are negligibly small -- especally relative to the big differences driving the results I find, e.g. that some registrars are preferred by domain warehousers and cybersquatters, others by folks who actually put their domains to use. Any good reason -- even just a casual story -- to think the contrary?