Re: [Uta] opportunistic keying / encryption considered of dubious value

[Keith Moore <[e-mail]>]

On 03/15/2014 02:43 PM, Watson Ladd wrote:
> All DANE/TOFU/latching imply a commitment by both sides to maintain 
> and manage identity. This cost is why we don't encrypt today.

I don't think so.  I think we don't encrypt today because many 
application protocols solidified in the late 1990s when there were more 
barriers to use of encryption, and there simply hasn't been sufficient 
pressure to update them and overcome the inertia of the installed base 
until (perhaps) now.

Also, I don't see how DANE implies a commitment by both sides.  And I 
don't think latching or TOFU requires (much) additional effort on the 
part of sysadmins.

> Furthermore, with the new email headers, we can probably design
> mechanisms to say "send only to nodes with verified identity. If you
> run into trouble don't." vs "hide me from casual inspection" vs. "it
> needs to get there".

I'd need to see a specific proposal, but offhand trying to embed such 
requests in email headers strikes me as poor design (i.e. layering 
violation) and also easy to defeat.

Keith