On 29 Oct 2013, at 19:46, Jesse Thompson <[email protected]> wrote:

> 
> 
> On 10/29/2013 12:59 PM, Dave Cridland wrote:
>> On Tue, Oct 29, 2013 at 5:46 PM, Peter Saint-Andre <[email protected]
>> <mailto:[email protected]>> wrote:
>> 
>>    -----BEGIN PGP SIGNED MESSAGE-----
>>    Hash: SHA1
>> 
>>    On 10/29/13 11:40 AM, Jesse Thompson wrote:
>>     > On 10/28/2013 2:52 PM, Peter Saint-Andre wrote:
>>     >> On 10/28/13 1:41 PM, Jesse Thompson wrote:
>>     >>> Are there more details?  Specifically, does "hop-by-hop
>>     >>> encryption using SSL/TLS" require strong association between a
>>     >>> domain name and an XML stream as described in
>>     >>> draft-ietf-xmpp-dna-04?
>>     >>
>>     >> We, as a community, need to figure out what we can do.
>>     >>
>>     >> Realistically, I think we need to prefer authenticated encryption
>>     >> via PKI, POSH, or DNSSEC/DANE and fall back to opportunistic
>>     >> encryption via TLS + dialback.
>>     >
>>     > So, the presumption is that servers which aren't capable of at
>>     > least TLS+dialback will be cut off?
>> 
>>    Yes.
>> 
>>    Now, this is a proposal, not an ultimatum. We, as a community, need to
>>    come to a decision about whether this is a reasonable course of
>>    action. However, I do think we owe it to the users of our services to
>>    provide a higher level of security.
>> 
>> 
>> Also, if phrased right, we could say that the Good Servers talk with
>> each other securely, but they may also have exceptions to deal with
>> legacy services which do not yet perform full security.
> 
> If being an exception is the past of least resistance - for both the operator 
> needing to change as well as the operator who is compelled to enforce the 
> change - then how do you prevent everyone from being an exception?
> 
> I like the proposal to "provide user or administrative interfaces showing 
> [TLS details]" because that has the potential to cause end-users to bug their 
> service operators to implement better security, which will cause service 
> operators to bug server developers to implement new security features.
> 
> That seems like something that can start phasing in right away.
> 
> Is it reasonable to expect the popular XMPP clients to begin showing TLS 
> information to end-users earlier in the proposed timeline?

On the topic of user-interfaces:

- How does a a server that fails to setup a s2s session indicate the failure 
back to a client?
- Does the protocol support an error message saying "certificate failure" or 
"TLS not available"?

/O

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to