ICANN/GNSO GNSO Email List Archives

[registrars]


<<< Chronological Index >>>    <<< Thread Index >>>

Re: [registrars] My comments - Verisign batch pool issue

  • To: "Bhavin Turakhia" <bhavin.t@xxxxxxxxxxxxxx>
  • Subject: Re: [registrars] My comments - Verisign batch pool issue
  • From: "Jordyn A. Buchanan" <jbuchanan@xxxxxxxxxxxx>
  • Date: Tue, 12 Oct 2004 17:31:35 -0400
  • Cc: Rusty Lewis <rlewis@xxxxxxxxxxxx>, "'Registrars Constituency'" <registrars@xxxxxxxx>, "PJ' 'Bolanos" <pj@xxxxxx>
  • In-reply-to: <200410102250.i9AMo1l17577@pechora.icann.org>
  • References: <200410102250.i9AMo1l17577@pechora.icann.org>
  • Sender: owner-registrars@xxxxxxxxxxxxxx

Bhavin:

There are several points in your e-mail that I take issue with. My thoughts are in-line below:

* the real problem is not verisign getting pounded. I don't think that ever was the issue. The number of connections were reduced from 40 per registrar to 10 per registrar in the last 8 months. There is statistical data from reliable that shows that there were more commands being sent to the batch pool in Jan 2004 than there are as of today. Verisign can clearly reduce the number of connections from 10 to 5 or even 1 and reduce their load by 1/40th
of the orignal load. The orignal system supported 100+ registrars at 40
connections. At 1 connection per registrar they should be able to support
4000 (I don't think we will reach that number ever considering the new
NSI/TUCOWS model as such discourages any phantom creds now)

There are several problems with this analysis.

You may not have been around at the time, but the reason that the automated batch pool exists is that the add storms in early 2001 crippled the registry for hours at a time and made it impossible to perform new registrations or make changes to existing ones. Given the opportunity, clearly have the ability to overwhelm the registry.

Second, as Bruce Tonkin has pointed out in a separate issue it is unacceptable for new entrants into the registrar space to simply dilute the resources available to existing businesses. Some registrars (Register.com included) need more than 1 connection to the automated batch pool simply to maintain our existing domain names. Long before the registry's theoretical capacity of 4000 connections spread over 4000 registrars is reached, a number of existing registrars will be unable to effectively serve their customers.

* Solution 1 also gives more chances to larger registrars, thus being
unequal

Several people have raised this issue, but I believe it has not been substantiated by anyone. What is the analysis here? Solution 1 rewards registrars that are efficient, regardless of size. Large registrars who simply pound away will exceed the ratio because they won't be adding names efficiently that way; small registrars who figure out a clever way to grab names with a smaller number of transactions will be able to continue to grab expiring names without

Solution 1: DO NOTHING
----------------------

* This is the simplest solution and it works. The market has considerably
changed since this debate came up.

* Firstly most registrars may shortly not allow names to expire judging NSI and TUCOWS' latest move. Infact I am quite certain of this eventuality - for various reasons which I will separately discuss. This has already prevented addtl phantom applications from applying. I will send out some hard data on
this shortly

The problem with this approach is that it seems to assume that the move that NSI and Tucows has taken is a good thing for the industry. I'm not sure I have enough information to really judge that yet. At the very least, their approach seems to present a number of problems, the most significant of which is that they are creating registration "islands" in which expiring names cease to be available to the general community of registrars. In the short term, that seems to be great for a company like Register.com, with a base of high quality names. From a slightly longer term perspective, though, I don't think we are stronger as an industry if registrants have to hunt around to figure out which registrar they need to go to in order to register a particular name. Indeed, average Internet users are unlikely to bother, meaning that expiring domains will continue to simply be re-allocated to businesses based around traffic generation. And while I recognize that traffic generation is a legitimate business, and many of us derive significant portions of our revenue from those who specialize in it, I think as an industry we should be looking for ways to make domain names valuable destinations in their own right.

In his articles about Tucows' approach, Ross admirably makes the case for including the existing registrant in the process in order to yield decisions based on "perfect information". Unfortunately, in the process sales channels are balkanized so new registrants visiting a particular registrar's site won't know that names being deleted by another registrar might be available at auction. The end result is that by moving closer to "perfect information" for incumbent registrants, we are moving farther away from that same idealized "perfect information" for potential new registrants. Given that the incumbent registrant doesn't care enough about the name to bother to renew it, I'm not sure this is a worthy trade off. Fortunately, even Ross seems to acknowledge that Tucows' local solution is only a transitional phase prior to some sort of global fix. We need to identify that fix before the market becomes fragmented.

Bruce Tonkin has already made the case that an auction seems like a reasonable way to resolve contention for names at the registry level. I agree. Such an approach allows the market for expiring names to operate without placing an undue burden on the registry and also has the potential to be profitable for a variety of registrars, including large registrars with an existing pool of names under management as well as smaller registrars who can now present a more attractive inventory of names to their customers.

Jordyn




<<< Chronological Index >>>    <<< Thread Index >>>