Re: [openpgp] Web Key Directory I-D -07

[azul <[e-mail]>]

Hello,

Am 14.11.18 um 11:09 schrieb Werner Koch:
> On Tue, 13 Nov 2018 22:37, bartbutler=[e-mail]
> said:
>
>> "The key MUST carry a User ID packet ([RFC4880]) containing the email address to which mail sent to the queried email address will be routed."
> You are talking about how mail is routed, the spec is about discovering
> the one and only key to be used for a given mail address.  And by key I
> mean the OpenPGP keyblock, that is the public key plus one user ID (or
> several if they have the same addrspec part).
>
> A mail address is here considered as an identifier for an entity and not
> as an addressing scheme for mails.  An entity may have several
> identifiers like Werner.Koch@foo, Werner_Koch@foo, wernerkoch@foo,
> wk@foo, koch@foo.  That is a pretty normal but there is no way a sender
> can decide whether they are all the same; for example the last two of
> the list could also identify my brother.
>
> A sender gets hold of one mail address and that must have been relayed
> (direct or indirectly) to them by the owner of that mail address.  The
> recipient needs to take care that a key exists for that very mail
> address.
I think the last sentence exactly captures the difference between the
two approaches.
Who needs to take care that a key exists / pick an existing key for a
different email address routed to the same mailbox.

One of the use cases I have for the '+' syntax is registering to
services with email addresses that allow me to remember where i used
them and filter mails based on them.

One example would be using '[e-mail]' to register
to a conference.
It's low overhead. I can just come up with a addition to my email
address when filling in a web form. If I later receive spam to that
address I know who leaked my email address.

Providing a key for the new address or even publish all those addresses
in my key do not seem like an option for me. Therefor the party I
submitted the form to will not be able to get back to me with an
encrypted email even if i have a key in the wkd.

I don't quite understand the downside of relaxing the spec in this
regard. As far as i understand it it would not require implementations
to lookup which mailbox an address would route to. It would just allow
it. For some providers / implementations this may be trivial and other
may choose to not do it.

Where does the need for the matching user_id come from?

One downside I see is that the mechanism could be used to detect where
an email address routes to. In the case of azul+conf this may be easy to
guess. But for aliases users do not expect them to be linkable without
any interaction on their side.

Cheers,
 Azul