[registrars] Verisign batch pool advisory

Bruce Tonkin Bruce.Tonkin at melbourneit.com.au
Thu Oct 7 04:53:10 UTC 2004


Hello Paul,

> Even if VeriSign spent nearly an infinite amount of money on 
> this problem (to "expand their capabilities"), and if we kept 
> the status quo, then, because it costs very close to 
> absolutely nothing to pound the crap out of the registry, all 
> registrars would increase their registry pounding rates to 
> the level that would immediately use up absolutely all the 
> vast capabilities
> that the nearly infinite amount of money purchased.   While 
> at the same
> time, we would not register even one more name than we did 
> with the system that did not have the vast capabilities.

Exactly.

Verisign really needs to develop an appropriate model in consultation
with registrars for allocating a recently deleted name when there is
more than one registrar seeking access to that name.  Ie this is
essentially a contention problem.   The current method of resolving
contention is ridiculous.   The first-come, first-served model is really
only appropriate where there is little chance of contention for a
particular name at a particular time.  If you were to announce on radio
(or the Internet) that the first person to be at your house by 9am
tomorrow would get a $1 million - I am sure you will appreciate that
first-come, first-served model would not work.  If however you were to
announce that the first person to be at your house by 9am tomorrow, will
get the chance to buy your old car at the current market price - then
that would be different.

The methods for resolving contention include:
- sealed bid
- auction
- lottery ticket
- choose a registrar at random from a list of registrars wanting the
name

.biz tried the lottery model and that presented problems - particularly
on an international level, as I assume various licences etc are required
to run a lottery.

Some registries tried variants of the random model when opening up their
name spaces for registration.

I believe the most economically efficient model is auction.

I am interested to hear if there are other methods (which scale).

If the auction model is chosen (whether to obtain a name when a new
registry starts, or when a name is released again for registration),
then the more interesting question is who gets the proceeds.

Regards,
Bruce Tonkin








More information about the registrars mailing list