Re: New proposal from Ronald J. Fitzherbert

[Simon Higgs <simon@higgs.com>]

At 10:35 PM -0800 12/1/96, Kent Crispin wrote:

> Perry E. Metzger allegedly said:
> > Kent Crispin writes:
> > > [commenting on the proposal for a 30 day waiting period for domain names]
> > >
> > > And it tremendously simplifies the issue of synchronization in shared
> > > domains, as well.
> >
> > Not at all. There will still be disputes as to who got to the domain
> > first even if it only gets installed thirty days later. It is still
> > necessary to have a true locking mechanism in place to assure that
> > such disputes do not arise.
>
> Authenticated email to a neutral third party announcing a claim to the
> name is sufficient -- the email's time of arrival at the neutral third
> party would be the "locking mechanism".  If the arrival time at the
> NTP was *exactly* the same, then a contractually binding random choice
> would be made by the NTP.  Thirty days gives plenty of time for
> "email-granularity" protocols to work.
>
> I suppose you could call this a "true locking mechanism" :-).
>
> As we both know, the locking problem is a very minor point -- even for
> very popular domains like .com the liklihood of close calls like this
> is really very small.  It is *possible* that I beat out somebody else
> by seconds for the name "songbird", but I really doubt it.
>

Sorry guys. 30 days for email is still a joke. It's December 1st and
I'm still receiving mail sent in July thanks to a couple of careless
mail servers out there (and no the TTL on our MX is not that long!).
I'd rather not have "locks" floating around the net in some queue
somewhere.

Use GPS on each server and create an NTP-based locking mechanism. The
arrival time of the "lock" packet is set by CPU ticks on the neutral
third party server. Make it only accept registration reservations
sequentially.



Simon

--
If at first you don't succeed, skydiving is not for you.