[TLS] Inconsistency/redundant length encoding of RenegotiationInfo structure

"Brian Smith" <brian@briansmith.org> Thu, 28 January 2010 04:30 UTC

From: Brian Smith <brian@briansmith.org>
To: tls@ietf.org
Date: Wed, 27 Jan 2010 22:30:18 -0600
Message-ID: <000501ca9fd2$9db45470$d91cfd50$@org>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Content-Language: en-us
thread-index: Acqf0pg+dkubdL9jRrepksMr2EMc8A==
Subject: [TLS] Inconsistency/redundant length encoding of RenegotiationInfo structure
Precedence: list

An "empty" RegenotiationInfo structure is encoded like this:

    0xff 0x01 extension_type
    0x00 0x01 extension_data length
    0x00      redundant length byte

Why does RenegotiationInfo needs this redundant length byte? It seems to me
that it adds nothing except extra code in the implementation--in particular,
more cases that have to be tested.

An empty RFC5077 session ticket extension is encoded like this:

    0x00 0x23 extension_type
    0x00 0x00 extension_data length

An empty server name extension is similarly encoded:

    0x00 0x00 extension_type
    0x00 0x00 extension_data length

The extensions from RFC4366--Server Name Indication, Truncated HMAC, Client
Certificate URL, Trusted CA Key, and Certificate Status extensions are all
encoded like the RFC5077 extension--that is, without a redundant length
byte. I think the RenegotiationInfo extension should be encoded similarly,
because it lets implementations reuse more already-tested parsing code.

To be clear, I'm not talking about just the case where RenegotiationInfo is
empty. The redundant length byte seems to be a useless complication in all
cases.

Regards,
Brian

[TLS] Inconsistency/redundant length encoding of … Brian Smith
Re: [TLS] Inconsistency/redundant length encoding… Martin Rex