Re: [TLS] Wrapping up cached info (and PRF WTF)

[Martin Rex <[e-mail]>]

Kemp, David P. wrote:
> 
> And on a different topic, What is This Function?
> 
> 
> RFC 5246 section 4.3 defines the notation for two types of vectors:
>   Fixed length:     T T'[n];
>   Variable length:  T T'<floor..ceiling>;
> 
> and section 7.4.9 defines the structure of the Finished message:
> 
>       struct {
>           opaque verify_data[verify_data_length];
>       } Finished;
> 
>       verify_data
>          PRF(master_secret, finished_label, Hash(handshake_messages))
>             [0..verify_data_length-1];
> 
> 
> What sort of notation is a function PRF() followed by a floor and
> ceiling expression enclosed in square brackets []?

The output of the PRF() is substantially larger than the size
of the finished message.  The notation is probbly meant to
refer to the initial verify_data_length octets from the
opaque result of PRF().

The "verify_data" vector in the finished message has a fixed size,
which means that it is encoded without a seperate length field.

But since the verify_data is also the only content of the finished
handshake message, the size of verify_data is the same as the
sized handshake message body given in the handshake message header.


> 
> Assuming it means that the PRF output is truncated to some variable
> number of bytes, why is the floor 0 bytes and the ceiling 11 bytes when
> all verify_data vectors MUST be either exactly 12 (current ciphersuites)
> or at least 12 (future ciphersuites) bytes???

The existing spec does not allow verify_data to be of variable size.
(variable size objects would require an additional length field
 preceding the value).

verify_data must always be fixed size, but that fixed size can be defined
in the specification of a cipher suite to be larger than 12 octets.

-Martin