Re: [apps-discuss] For consideration as an appsawg document: draft-hoffman-server-has-tls-03.txt

[Paul Hoffman <[e-mail]>]

On 1/25/11 11:11 PM, Patrik Fältström wrote:
> I have had a look at this, and have three questions:
>
> 1. Have I understood it correct if the client by using this RR can
> refuse to fall back to insecure connection for "communication using
> HTTP"? I.e. it seems the RR is specifying whether fallback is
> possible to not only per port, which works when you have "STARTTLS"
> or similar, but not when fallback is between ports?

The intention is to tell a client that could fall back, regardless of if 
it on the same port (STARTTLS) or between ports (443 to 80), that the 
server offers the non-secured option. I can try to make this clearer in 
a future draft.

> 2. If so, the "input" to the query should in that case be not only
> the hostname, but the protocol and hostname, right? So one could use
> a prefix-based mechanism like _http._tcp.example.com. IN HASTLS 0 443
> 0

That might be good to do anyway. That is, having the query be in the 
same format as SRV queries, but with a HASTLS type, could give just the 
information wanted. The result could be an RRset with multiple HASTLS 
records.

> 3. I am always a bit nervous over RDATA that has variable length. Do
> you have an implementation of this so that you have tried to ensure
> "it works"?

No, and I think that the above proposal might allay your fears.

> Otherwise, as someone said, overall negotiation I think is
> interesting.

Great!