Re: [TLS] Last Call: draft-ietf-tls-rfc4366-bis (Transport Layer

["Blumenthal, Uri" <[e-mail]>]

I support option (2) and am against option(1), for reasons expressed in earlier emails.

Tnx!


----- Original Message -----
From: Simon Josefsson <[e-mail]>
To: [e-mail] <[e-mail]>
Cc: Blumenthal, Uri; [e-mail] <[e-mail]>
Sent: Thu Oct 01 07:00:34 2009
Subject: Re: Last Call: draft-ietf-tls-rfc4366-bis (Transport Layer

Martin Rex <[e-mail]> writes:

> Blumenthal, Uri wrote:
>> 
>> In my understanding, TLS-established server_name should be
>> enforced by the server.
>> 
>> And Martin - I couldn't disagree more with you. The whole point of
>> using TLS is to enforce who can access what. So the client makes sure
>> he accesses the right server, the server makes sure he grants access
>> to the right pages on the right virtual host. And if your server
>> doesn't do that - please kindly tell me what commercial or
>> freeware product it is included in, so I can avoid buying or
>> using it in the future.
>
> There seems to be a significant misunderstanding.
>
> The Host header field of an HTTP request is a detail of an
> application protocol.  The hostname conveyed by the TLS extension
> server name indication (SNI) happens at a competely different
> protocol layer.
>
> The difference becomes obvious when you add reverse proxies
> into the picture (those which terminate the TLS wrapping).
>
> Conceptually, the Host: header field of a HTTP request is
> part of the URL.  If a reverse proxy perform URL rewriting,
> it may as well have to rewrite Host: header fields.  That
> depends entirely on the backend architecture of each
> particular software installation.
>
>
> Whether or not an application may want to make consistency
> checks between a Host: Header field and a hostname received
> via SNI at the specific point of the backend architecture
> where TLS was terminated depends entirely on the backend
> architecture, and is an application issue.

I believe this wording in RFC 4366bis makes it a TLS issue:

   If the server_name is established in the TLS session handshake, the
   client SHOULD NOT attempt to request a different server name at the
   application layer.

I believe there are two options:

1) Either remove all requirements on application behaviour from 4366bis
   (including the text above) and explicitly defer such discussions to
   other documents.

2) Add the text I proposed to make servers actually validate proper
   client behaviour.

I went for 2) assuming that the text above was intentional, but I share
your arguments for going with approach 1).

/Simon