On 05/17/2014 11:13 PM, Robert J. Hansen wrote:
> A lot of users would object.  "I don't want Enigmail to keep track
> of whom I'm corresponding with and what technologies they're using!
> It's a violation of privacy!  How do I know what you're doing with
> that information you're collecting?"
> 
> And I hasten to add -- it's a pretty reasonable objection.

It's absolutely not a reasonable objection. If you really want to
appease those users, then lobby on the list for an option to disable
that information collection, but don't lobby against measures to
increase convenience and (thus) security for rational users in order
to placate fools. IINM, PGP/MIME masks more metadata than the other
options, and so it would be preferable for more mail to be sent using
that, from a privacy and security perspective.

OTOH, I can't conceive of any security risks imposed by maintaining
such a simple list of contacts. In my proposal, it's just a simple
blacklist, a list of contacts, nothing more; not even the particular
clients seen being used by the contacts. I suppose that you could
argue that in the case of a user deleting all local copies of messages
from a contact, you retain an additional record of the user having
communicated with the address of the contact if they're in the black
list, and so if the user's machine were apprehended and had forensic
analysis performed on it, you could say that that's a potential
problem, the additional data retention. But I'm willing to bet that
Thunderbird's own "collected addresses" functionality does that already.

> Just to be clear: I don't know of any MUA that can handle/decrypt 
> PGP-encrypted mail but *can't* handle/decrypt PGP/MIME-structured 
> encrypted mail.  All the PGP/MIME problems i've seen reported have
> to do with failing to read PGP/MIME-signed (but not encrypted)
> messages.

Oh, I didn't know that, but it's interesting, in that it would allow
for another heuristic to be added in terms of Enigmail being more
proactive about deciding the encryption scheme to use: if the message
isn't set to be signed anyway, then there's simply _zero_ reason not
to go with PGP/MIME, right? That cuts out a whole bunch of cases right
off the bat.. thoughts, anyone?


> * just because you've never been sent an e-mail from that user via
> a PGP/MIME-broken MUA doesn't necessarily mean that they never read
> mail on such a MUA.  Sending is not receiving.

Okay, so how do you personally determine whether or not to send
someone a message in PGP/MIME, then? What it does mean is that the
user will at least have some way to decrypt and read the message on
one of their devices, somehow, without needing to install and
configure any new software. If, given the recipient's personal mail
routine, it's a huge nuisance, the user could respond requesting that
the sender make a rule for them that unsets the PGP/MIME option for
future emails.

In any case, IINM, automatedly or not, it's the only data we have to
go on, save for exceptional cases like, say, people who live together
and incidentally see each others' screens and know what each others'
setups are or something like that.




> * you'd need to enumerate which versions of which MUAs (with which 
> plugins) are PGP/MIME-broken and keep that list up-to-date somehow.
> How does that interact with, say, webmail implementations?

I'm not terribly well informed, but I'm under the impression that it's
only a few stubborn versions of outlook that choke on PGP/MIME, and
that the latest versions of MS' MUAs even do fine with it, so I don't
think it would need to be a dynamically maintained list that would
ever get any additions to it. I can't imagine that any new mail
encryption packages in the future would not put the effort in to
support the obvious optimal encryption scheme (PGP/MIME). This strikes
me as being like saying that you'd need to enumerate which browsers
don't support addEventListener so you can serve them a version of
pages that use attachEvent instead.


> * some messages don't have any clear indication of what MUA they
> used.

That's a point.

> * someone who used a PGP/MIME-broken MUA 6 years ago may not still
> be using that MUA.  should six-year-old mail still be relevant?
> Figuring out what sort of cutoffs are reasonable is non-trivial.

I'd say that if the user ever sent mail indicating a PGP/MIME-enabled
MUA, then they should be put on a white list, because it means that
they have the means to read mail. If for whatever reason, that's not
the way they'd like to get mail in the future, they can request to
have a rule set for them, and if there was a problem with that initial
message, they can request a resend.

Ultimately, I'm pretty biased, because it seems to me that PGP/MIME
conceals more metadata, and metadata collection is a collective issue.
Both sender and receiver are stakeholders, and ultimately, even people
who aren't a party to the communication are, too, because the
government makes inferences at the edges of the social network maps
that it builds.

When you're not the only stakeholder in a matter, it's sort of bad
etiquette to be using software which doesn't support modern standards.
Much like the fact that old-IE's prevalence is a real social problem
makes the use of old-IE socially irresponsible, so too, I think, does
the fact that mass government surveillance is a social problem make
using outdated encryption software rude.

As developers of a prevalent mail encryption package, we have the
power to give significantly consequential nudges of the GPG user base
towards a critical mass of modern mail encryption standards supporting
software adoption, and I think we have a social and political
responsibility to do that. In that regard, the points of inconvenience
you mention are actually a feature, not a bug.

> * Imagine a user who has 300,000 message in their mail archive.
> You wouldn't want search all of those each time.  so you'd want to
> keep some sort of state for each user that could be updated when
> new mail from that user was discovered by thunderbird.

That's what I had in mind.

> This doesn't necessarily just mean when new mail is fetched --
> thunderbird could suddenly get access to an old mailbox, or a new
> folder could be published.  how would you synchronize this cached
> state?

Not sure what you mean by this, on syncing cached state. Sorry.


Anyway, with the system in place now, what thought process does
sending user Bob go through in his head when he's presented the modal
scheme selection prompt? "Hmm, gee... I guess I remember reading
somewhere that PGP/MIME is ideal if the user's mail client can render
it, but.. wait, is that for decryption, or only processing signed
mail, or... oh, that's right. It isn't a nebulous unenumerated set of
clients that don't accept PGP/MIME, it's just some versions of one or
more of the MS MUAs. [So what the question of 'should I use PGP/MIME
for this message or not?' really comes down to is 'is my recipient
prone to read mail on a MS MUA?', and it's quite uncommon for people
to be familiar with their email contacts' software diets, and so the
only really reasonable way to respond to that modal, even as things
are today, is by using the data at your disposal, i.e. the headers of
the messages you've previously received from that contact. But that's
unreasonably tedious, and so I'm proposing that we simply automate the
only rationally-grounded way of answering that modal prompt.] Oh well,
I guess better safe than sorry, and I don't want to give my contact
friction when opening my message, so let's nix the PGP/MIME." That's
unfortunate, because it results in a more surveilled world for all of us.

Cheers :)

_______________________________________________
enigmail-users mailing list
[email protected]
https://admin.hostpoint.ch/mailman/listinfo/enigmail-users_enigmail.net

Reply via email to