Re: [dane] draft-wouters-dane-openpgp-01 review

[Jelte Jansen <[e-mail]>]

On 07-01-14 07:32, Mark Andrews wrote:
>
> All of which is irrelevent provided you can encode that into a policy
> which can be transmitted.
>
>> It is not possible to handle these without substantially complicating
>> the logic.  One would have to query the domain for the domain's
>> recipient delimiter first, and then for the address.
>
> So.  One of the reasons to go with base32 and not raw binary is
> that the DNS does normalisation which is potentially different to
> the normalisation done by the SMTP server.
>
> At a minimum we should be able to specifying "no normalisation" vs
> "case fold" (and which direction) for ascii LHS.
>

Seems to me these are both the same issue, and they could either be 
resolved with one (probably nasty) policy-specification, or not at all 
(leave 'em out like smtp does). Not a big fan of either :)

> Yes, it makes things more complicated but the real world is
> complicated.
>
> Remember that one is comparing this to a SRV record which points
> to a key server that does all the normalisation required to return
> the correct key 100% of the time.
>

Are you suggesting this as a possible solution (which sounds more like 
using DNS(SEC) to specify personal pgp key servers to get around the 
current pgp key server problem)?

(while I type this, I wonder whether this functionality should be on 
this level in the first place; shouldn't it fit better in smtp itself, 
which is the only real place that currently know what normalization and 
other rules apply?)

One small thing on the draft itself: IMO the last part of section 2 
should not use 2119 terminology; it's not about interoperability nor 
implementation. Oh and 'sent' should be 'send' :)

Jelte