[nc-whois] some questions and remarks

[Thomas Roessler <roessler@does-not-exist.org>]

I should probably start with a word of warning: I'm a newbie to this 
task force, so, in particular, I don't know what has been spoken and 
discussed during earlier telephone conferences.  All I know is what 
I could gather from the list archives, from e-mail conversations, 
 from Monday's teleconference, and from the material which was just 
distributed off-list.

Still, some questions and remarks remain, and answers from those who 
have been present on this task force for a longer time than myself 
would be most appreciated.



1. I don't seem to be able to locate the original set of questions 
asked in the survey - the form has disappeared from ICANN's web 
site.  Is the assumption correct that the survey results distributed 
off-list contain the original wording of questions?



2. I don't seem to be able to locate this task force's terms of 
reference (maybe I didn't dig deep enough).  From Monday's telephone 
conference, I seem to understand that the main question we are 
trying to answer is: Should the WHOIS service be changed?  If so, 
how?  Additionally, it seems that, thanks to a recent ICANN board 
referral to the DNSO, we also have some more explicit questions 
concerning bulk access to whois data to answer. 
<http://www.dnso.org/clubpublic/council/Arc09/msg00061.html>

Is this correct?



3. Is the impression correct that no indermediate questions we are 
seeking to answer when reading the survey results and responses have 
been defined so far?



4. There is what looks like a contradiction between the answers to 
q16 and q17 (bulk access; just look at the statistical results). 
 From Monday's telephone conference, it seems that at least Steve 
Metalitz and I have strongly diverging views on how to interpret and 
resolve this.  Can we please add this point to the agenda of one of 
the next telephone conferences, or discuss it on this mailing list?

If this has been discussed before: What's the result?



5. Concerning categories of respondents.  For some questions, it may 
be a good idea not to break down the answers by the category of 
respondent, but by the primary use made of the whois system (in 
particular, that's the question of adequacy of the whois system for 
respondent's purposes - for instance, "government" respondents may 
include system adminsitrators from government research labs as well 
as law enforcement).  We may wish to look for more specific profiles 
("techie", "IP lawyer/law enforcement", "personal home page owner"), 
which may turn out to be partially orthogonal to the categories from 
question 1.

In fact, I believe that we have a strong methodological problem with 
this survey: We seem to have no distinction whatsoever between 
respondents who are answering in their capacity as whois system 
users, and between respondents who are answering in their capacity 
of being registrants, i.e., those whose data are stored in the 
system.  There are some questions where it may be apparent in what 
capacity people answer, and there are some questions which provide 
us with hints on who's answering the survey form.  Still, this 
doesn't make things a whole lot clearer.

We may wish to discuss if we can formulate some machine-evaluable 
criteria for different classes of respondents.  For instance, 

((is not in categories ISP or registrar-registry) AND
 (has registered a domain name)  AND 
 ((uses whois never or occasionally) OR
  ((has registered few domain names) AND
    (mainly uses whois to check for availability))))

may be a check used to single out registrants.  If we can agree on 
such criteria, we may wish to approach ICANN, so they can apply 
these criteria to the original set of data (I suppose that the 
original data are stored in some SQL database, so this would 
probably be a single SELECT statement).  I don't believe there's any 
value in trying to do this manually, when the data is (somewhere) 
present in machine-workable form anyway.



6.  Concerning "baskets", I believe we should distinguish two kinds 
of "free form questions": (1) multiple choice gone wrong, and (2) 
asking for creative writing.  

Class (1) certainly covers (some? all?) the "other" responses to 
multiple choice questions, q14, the cost issue of q10 (ultimately, 
it's either "registrants" or "users of extended search service", so 
thi should have been multiple-choice; note, by the way, that the 
implementation side of this is entangled with the bulk access 
problem in a most interesting way), and others. For this class of 
questions, a "basket"-based approach would certainly reasonable.


Class (2), however, is different: This includes the question of what 
kind of harm happened to registrants (BTW, the "yes/no" answer texts 
given in the material mailed out today are problematic in this case 
 - after all, whois users may have experienced inaccurate data 
without experiencing measurable harm or inconvenience).  This 
includes most of q20.  In fact, I believe that we should be careful 
about restricting ourselves to a "basket"-based approach with this 
class of questions.  It may be a better idea to look for answers we 
find surprising or interesting (like Oscar did), and which do not 
just confirm what we believe to know from the statistical data). 
I'd suggest that we try to collect a finite set of such questions or 
ideas, and discuss these by their merits, using the statistical 
answers as background for this discussion.

Thus, before we start to agree on baskets, we should most likely try 
to agree on what questions we can resaonably access at all using 
that kind of approach.



Kind regards,
-- 
Thomas Roessler                        http://log.does-not-exist.org/