>>Okay, well, how should they solve it, then? If you have an idea, I'd love
>>to hear it (and I'm not being sarcastic). I honestly don't think it's 
>>solvable -- there is a virtually infinite demand for some names, and the 
>>registry/registrars/resellers can't supply virtually infinite connection 
>>resources.

HW> Simple. Every registrar has access to the batch-pool.  The registry
HW> complains of tens of millions of worthless "checks" during drop times.  An
HW> additional response code signifying a name has been registered IN THE LAST
HW> 24 HOURS would stop the registrars from continuing to pursue that name.

YES!  I believe Tucows proposed a very similar idea to Verisign:
Right now, if you do a "check" command on a domain, and you are not
the registrar for that domain, you receive an error message, otherwise
you get the expiration date.  The proposal was to have the check
command return the expiration date regardless of who was the
registrar.  This would eliminate *huge* number of check/add commands and
largely solve the problem, just as your IN_THE_LAST_24_HOURS flag
would.  Either one is fine by me.

There was another proposal which would also take a huge load off of
the registry: The top registrars do millions of check commands against
the registry on a daily basis to support normal operations. On average
there are less than 50K adds and 50K drops performed at the registry
each day. Let's say the top 4 registrars each do 2.5 million checks a
day (my guess is that this is a conservative number) for a total of 10
million checks a day. The idea is that the registry can push those
100K changes out to the registrars, thus saving 9.6 *million*
transactions a day, which is more than 100 per second, and when you
figure in the peaks, you may be talking 200 per second during busy
times.

I also think that the registry should publish a list of names to be
dropped and when they will be dropped to all the registrars.  This
would eliminate check/add commands on names that people think are
going to drop, but really aren't going to drop.

Implementing those 3 proposals (the first one could probably be
implemented in an hour) would probably bring the load on the registry
down to record lows.

regards,
-joe

Reply via email to