Re: [TLS] NULL cipher to become a MUST NOT in UTA BCP

[Viktor Dukhovni <[e-mail]>]

On Fri, Sep 05, 2014 at 03:30:55PM +0200, Bodo Moeller wrote:

> If NULL encryption cipher suites can't be enabled together with non-NULL
> encryption cipher suites, you'd break any server that attempts to allow
> NULL encryption for clients that decide they don't need encryption (e.g.,
> because they're connecting to localhost) while supporting non-NULL ciphers
> for clients that don't offer NULL encryption for whatever reason (e.g.,
> because the implementor just doesn't like NULL encryption, which I think is
> a position worthy of support).  Same for clients.  I'd rather prefer for
> the BCP to avoid this quagmire entirely.

This is why my suggestion was specifically about the initiator
(client):

    What is much more reasonable is that initiators (clients) should
    refuse to negotiate (should suppress any configured) NULL cipher
    suites when any non NULL cipher suite is enabled.  This prevents
    accidents, without breaking carefully considered NULL to NULL
    deployments.

It leaves room for dual-purpose servers.

-- 
	Viktor.