[Mailman-Developers] full anonymisation approach

[Stephen J. Turnbull]

Rashi Karanpuria writes:

 > > Aside: I think if you actually use such a list you'll discover that
 > > most "shy" people are far more afraid of being flamed than they are of
 > > social stigma *outside* of the discussion thread.  OTOH, it is likely
 > > to bring out the worst in trolls.  IMHO YMMV etc.

 > We can't overlook the fact here that shy people and most of the
 > others facing one social stigma or other do benefit from anonymous
 > conversations and 12-step program lays focus on such
 > websites.

I'm not overlooking the possibility, and if you want to use the word
"fact" you should cite serious research (or reliable textbooks that
cite the serious research).  Psychology (especially clinical) is a
field that's rife with real crap (see Kahnemann _Thinking Fast and
Slow_ for an introduction).  For GSoC purposes, I'm willing to assume
that there are benefits for the use case you describe.

What you're overlooking is that you use words like "anonymous" and
"conversation" without specifying them properly.  Is an "anonymous"
mailing anonymous enough for the users?  How about the FUD effect if
there is a public incident where a stalker manages to identify a
victim through your anonymous list implementation?  The design of your
implementation is going to depend sensitively on the answers you give
to these questions.

 > I plan on keeping list identity of a poster constant in a thread,
 > i.e., I generate a new list identity for a first post to a new
 > thread.  Reasons of this were mentioned by me in
 > http://www.mail-archive.com/mailman-developers%40python.org/msg15283.html

OK.  Remember that identifying threads is inaccurate.

I wonder if it would be useful to have a feature where a user could
change her ID, or link an old ID to a current one.  (That's not a
requirement for this product, but IMHO you should consider the
possibility with designing modules/internal APIs.)


 > So, we will be using the In-Reply-To header of the mail to map the
 > messages to their respective threads.  Hence tweaking with subject
 > will not be treated as a new thread as long as the user is using
 > Reply-To feature, which is essential while conversating in a
 > thread. Also using references won't be reliable as after the length
 > of thread becomes large few entries from the references are dropped

I disagree.  You should use both.  The RFC for References does not
specify which, if any, References are to be kept, but most
implementations keep all, most of the others I've seen keep the most
recent.  If any are kept, they help verify the members of the
predecessor set, even if the order can't be fully relied on.

 > > Speaking of "originator", what do you propose to do about
 > > non-subscribers who are CC'd?
 > 
 > We can remove the CC and BCC fields as if sent to a non-subscriber
 > using CC the original id of the originator will be displayed and if
 > a subscriber is CC'd then the mail bounces as posting address is a
 > fake address.

I don't understand what you mean by "posting address is a fake
address."

 > The originator can use one of the many available anonymous mailer
 > services for personal anonymous mails.

You absolutely cannot rely on originators to be clueful about this (in
fact, isn't that your original motivation: the obvious answer is "if
you want to be anonymous on a list, use an anonymizer service").