Filtering indirection [Was: Re: list moderation]

[Alexandre François Garreau]

Le mardi 5 novembre 2019 18:17:16 CET, vous avez écrit :
> Samuel Thibault writes:
> > Wow, this is so welcoming a community...
> > 
> > Samuel

I initially thought that was in answer to a message of mine, because my MUA 
placed it under…

> Ruben has been placed under moderation and I rejected the message that
> you are referring to.  If you received it, it's because he sent it to
> you personally (I guess by scraping the email addresses of everyone who
> has participated).  I will see what options are available to us.

That raises two problems, a social and a technical one.

The second, technical, for censorship failure, is that to solve this problem 
currently involves better knowledge from users.  That can be expected of 
Samuel… but will everybody know it?  We must find a (likely technical, at 
least) solution, as if people won’t learn how to personally block/filter 
someone (possibly easily, collectively, or maybe even automatically), maybe 
one day people will argue something such as “develop international procedures 
so to force mail servers censor mail or shut down mail servers by force or 
ignoring from other lawful mail servers” that would break the internet we 
know.  Imagine gmail doing that.

For instance, I first thought “but… you should not say *that* about *that* mail 
if it didn’t got the List-Id/List-Post!”, but that’s too technical (it’s a 
problem most MUA currently don’t treat specially by default list messages), 
and then “MUA shouldn’t include the list address if there’s not the lists 
headers” …but that’s negative reaction: wouldn’t work (for instance for anyone 
not subscribed and never having received mail from the list (hence ignoring 
it’s a list)).

Actually a censoring list in a Cc would act exactely as someone who still 
receive mail from a thread person started (or was tiercely put in copy), but 
doesn’t want to hear about: so you can’t be expected not to answer because of 
it… But actually, that’s encouraging: solving it would then solve the bigger 
problem!

So, normally, what is done? either person actively states to other recipients 
they don’t want to be included anymore (we’d need a new standard for doing 
that automatically when filtering occurs, like a formally defined header format 
for a control message to be sent), either person stop reading the answer per 
the same filter used before (either “refusing to read”, “recognizing as spam”, 
“not looking at/putting in the folder” (for instance when someone stops 
reading a whole mailing-list (that sometimes occurs sadly, though most often 
because of (globally relevant) mail volume, sometimes also for social reasons 
people here would like to avoid)) that is in user (interface) language: “fold 
the whole thread and stop reading it”.

To the former, that clearly means “formalize censorship rules” and maybe even 
“publish and update them” so to move implementation of it more local or even 
client-side (that includes to publicly state if moderation is pre- or post-, 
so to possibly tell people “(don’t) wait for particular per-message censorship 
rejection notice to read it”, about who, what, to what extent/time/duration/
end, etc.).  Then, when mailing-list receive a mail it filters, it answers with 
the mail refusal control message (that should look like a bounce) not only the 
author, but *all* participants (informing: “please don’t talk about it to us 
anymore, don’t put us in copy”, solving the sad situation that happens on IRC 
when you /ignore someone, and then see half conversations and possibly end 
recursively ignoring everybody).

To the later, that means indirecting the filtering rules by the “references” 
header (collect message-id’s of everything filtered, and filter the same when 
it 
references it (depending on the filter and configuration (you might want to be 
aware about the difference between what come from the list and what’s 
personal))): so the list could have filtered Samuel’s message (and other list 
participants wouldn’t have seen any quote or comment, even more especially 
under the wrong message).  And people having received the censorship control 
message could too.  I proposed that feature to mailman hackers some years ago, 
when similar censorship problems happened elsewhere (I think it was on a La 
Quadrature du Net (EFF french equivalent) ML), but they said it wasn’t a 
priority, code was long to hack, they didn’t have time, and I ought to do it 
myself, except it’s written in Python and I don’t write python (and it isn’t 
extensible with C or Guile (sad dream of “I can extend any software with the 
favorite script language I want” that never happened), nor will because this 
is python, not C or lisp).

Or the list could don’t filter anything, and people would, per-config, choose 
to 
see or not what moderators might have selected as “irrelevant” or “nasty” 
(those could even be categorized differently in different filter lists, and 
other 
people could publish other lists) and messages referencing these.  That’s my 
dream.  Censorship wouldn’t exist anymore except in *explicitly* authoritarian 
circumstances, and we could argue against them, and actively make anti-
censorship P2P systems that go against this automatically without added noise.