>>>>> "John" == John R Levine <johnl@xxxxxxxxx> writes:
>> I also think you have higher trust in dnssec-validated keys than
>> a key that you get from a key server without a trust path to some
>> key you trust.
John> Not really. The only credible assertion a self-signature like
John> DNSSEC can make is "this is me." But in this case it's "this
John> is my user" which is not the same thing. Unless you know
John> something about the relationship between the domain and its
John> mail users, that might be anywhere from completely true to
John> completely false.
Let's focus on the case where it's completely false, yet it's still
reasonable to trust the domain to publish the right MX records. I'm not
seeing that case at all, so I'd appreciate some help.
John> If it's false, it doesn't have to be false for malicious
John> reasons. If I ran a webmail service, call it GooHoo, I'd
John> publish keys for all my users. Why not? The opportunistic
John> encryption keeps random strangers from snooping on incoming
John> mail, webmail pretty much requires that the mail service
John> handle the encryption (there are plugins, but I've never found
John> one that was at all usable), and I can continue to enhance the
John> experience of my webmail users by displaying relevant ads from
John> our trusted marketing partners.
Is this supposed to be a case where it's completely false?
If so, I'm not seeing it.
I think it would be reasonable for a webmail service to generate a key
at account creation time and to publish these records for that key.
I'm not seeing the case where it's likely to be completely false for
malicious or other reasons. I mean, yes, yoru DNS operator could mount
a DOS on your mailusers by publishing bogus PGP key records. Why would
their incentives be aligned with doing that in cases where you're
willing to trust them enough to publish MX and other mail related
records?