[gnso-rds-pdp-wg] The meaning of the things we're talking about (was Re: Principle on Proportionality for "Thin Data"access)

[Andrew Sullivan]

Hi,

On Wed, May 31, 2017 at 10:09:46AM -0700, Michael Peddemors wrote:
> 
> We aren't discussing DNS or any other places that data is available as part
> of this working group. Only the informed consent of data held in whois thin
> data.

I feel like maybe some of us are talking at cross purposes because we
seem not to be talking in a common language.  I suspect that I have
some pretty dark corners in my understanding of the legal and policy
language here.  Equally, it seems to me that some others have a very
different understanding of some of these terms than the one I have.
Mine is informed by data theory and many years of experience with
these kinds of systems.  I thought maybe it'd be useful for me to lay
out what I understand, in the hopes that maybe others can spot ways in
which my understanding does not match their model.

In my view of the world, the fundamental purpose of the entire
registration system is to create name spaces on the Internet using the
DNS.  There are certain data that are absolutely necessary for that:
the name itself, a record of who is in control of it, the NS records
that create the delegation, and so on.  These things carve out a name
space from the global Internet name space (which is founded in the
root zone, which is why ICANN is involved in this at all).  People
might want to create domain names for reasons _other_ than making them
work via the DNS, but those are derivative and secondary purposes that
supervene on the basic purpose of the registration system.

When you create an entry in the domain name space (anywhere in it --
immediately beneath the root, beneath a TLD, or at
very.deep.delegation.crankycanuck.ca), you are performing a
"registration", and the authority who has the ability to create that
entry is called the "registry".  Over time, in the part of the name
space near the root, we created some formalisms about this, including
a model that looks a little like "wholesale" vs "retail" markets.  The
result of this is the R/R/R model that governs the contracted-party
world at ICANN.  The result of this is that we have a distributed
database, operated by multiple parties along the lines of separation
implied by the business model and the authority distributions.

Modern registries contain additional associated data about domain
names created in the bailiwick of the registry.  That is "registration
data", and the RDS we are talking about is the generic access protocol
by which that data may be accessed.

>From my point of view, the data "in the registry", "in the whois", and
(when relevant) "in the DNS" is _the same_ data.  It might be
instantiated in different databases -- that is, there might be
different servers where it lives.  But it's the same data.
Distinctions between what is "in the registry", "in the registrar",
and "in the whois" are meaningless to me for that reason.

Some additional data about the registrations are a result of the act
of registration itself.  Some of that data is held by the operator of
the registry, and some by the agents that perform registrations (the
registrars).  These are things like the creation and update dates --
in effect, metadata about state transformations and so on.  My
understanding is that, when we talk about "thin data", it is either
this sort of action-generated metadata, or else it is data that is
intrinsically necessary for using the data (e.g. to make names work or
else to make the RDS itself work).  It is that basic fact of what the
data is for that makes me think thin data cannot be subject to
controls for privacy and other such reasons.

Some additional data about the registrations are data that need to be
provided by someone.  This additional data is what I think we are
talking about generically when we refer to "thick data".  Thick data
has many species, and probably we are going to need to discuss in
depth the different kinds of data in there.

I hope this makes at least one perspective a little clearer.

Best regards,

A

-- 
Andrew Sullivan
ajs at anvilwalrusden.com