Re: [keyassure] Objective: Restrictive versus Supplementary Models

[James Cloos <[e-mail]>]

>>>>> "ER" == Eric Rescorla <[e-mail]> writes:

ER> 1. Protecting against CA mis-issue (as in CA/cert lock).

I hadn't though of doing this as Eric later describes (such that dnssec
isn't required for this usage).

I can understand how even an unsigned specification that a resource
uses this-and-only-this CA might be an improvement over the status quo.

But isn't false RR injection combined with a matching false cert chain
still a relevant problem in that scenario?

Perhaps the rfc should spec that, but with a threat analysis.

ER> 2. Allowing TLS to work without getting a certificate from one
ER>    of the established trust anchors.

The above said, this is my primary motivation.  Commercial trust anchors
provide little value and significant burden.  

It isn't just about "convenience".

Obviously there is still a lot of third-party trust involved.  You have
to trust that the DS RRs are properly handled by the registrars, root &
tld zone operators, et al.

But we already have to trust them to do the same with the NS glue RRs.

Ie, if the NS glue is mishandled the game is over anyway.  So why not
trust them correctly to handle the DS glue, too?

The one thing that does become significantly more convenient, though,
is more frequent rollover.

Eric's second scenario should be DANE's primary mission.

-JimC
-- 
James Cloos <[e-mail]>         OpenPGP: 1024D/ED7DAEA6