On Sun, 26 Apr 2015, Viktor Dukhovni wrote:
On Sun, Apr 26, 2015 at 04:59:12PM -0400, Paul Wouters wrote:
Great, it looks like the proposed standard for hardening SMTP/TLS
could be repurposed for either http(s) or arbitrary ports as per my
proposal no?
There is nothing left to harden. The presence of TLSA means, never go
to the insecure port.
Yes, when the client is not already committed to using TLS, i.e. it is
opportunistic.
Unfortunately, the use of the standalone word "opportunistic" is
confusing as it means different things.
I tried to get this meaing into the original TLSA spec, and there was
resistence to it. It was sidetracked into the HASTLS record, which never
saw the light. I'm not sure if the DANE OPS (SRV) draft clarifies this,
but any sane client implementation of TLSA should really assume this.
We don't currently have a generic "opportunistic DANE TLS" document,
the SMTP draft directly specifies this behaviour for SMTP. The
SRV draft mostly does the same, withouth explicitly calling this
out as being opportunistic.
I expect the OPS draft LC to start any day now. If it should
generalize this observation, this is a good time to suggest suitable
language.
I'm behind on my dane and dnsop document reading :(
When I started work on what is now the SMTP draft, it originally
was a generic opportunistic DANE TLS document. It later split into
SMTP and OPS, with the former not covering non-SMTP use-cases, and
the latter not covering opportunistic security.
The opportune part is "hey, they are publishing a key to use for
crypto". Once you're at that stage, doing TLS is not optional, but
mandatory (IMHO, because people did not want to commit to this
in the original DANE RFC)
Paul
_______________________________________________
dane mailing list
dane@ietf.org
https://www.ietf.org/mailman/listinfo/dane
