Re: [TLS] Comments/Questions on draft-gutmann-tls-encrypt-then-mac-00.txt

["Paul Bakker" <[e-mail]>]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of
Peter
> Gutmann
> Sent: donderdag 26 september 2013 6:55
> To: <[e-mail]>
> Subject: Re: [TLS] Comments/Questions on draft-gutmann-tls-encrypt-then-
> mac-00.txt
> 
> Martin Rex <[e-mail]> writes:
> 
> >When changing from mac-pad-encrypt to encrypt-mac, as suggested by
> >Peter's draft, the MAC no longer operates on the TLSComressed PDU, but
> >on the TLSCiphertext PDU -- which results in:
> >
> >       HMAC_hash(MAC_write_secret, seq_num + TLSCiphertext.type +
> >                     TLSCiphertext.version + TLSCiphertext.length +
> >                     TLSCiphertext.fragment));
> 
> So all that's necessary is to change the text to reflect this, and perhaps
a note
> to point it out?  That's fairly simple to do.

I'm rather new to the list and late for this party..

But 'skipping' a layer (the TLSCompressed one in this case) is a bad idea
for a number of reasons.

Code path verification.
In case this comes to fruition, instead of the clean:
  "Assemble Data -> Compress -> Cipher -> Send",
 we get
  "Assemble Data -> Maybe hash in some specific cases -> Compress -> Encrypt
and sometimes hash in some specific cases -> Send"

This makes checking code correctness a step harder, aside from the fact that
suddenly cryptographic information (which used to be only required in the
Cipher layer) suddenly needs to be available on other points as well..
Meaning more coupling between layers and functions. This is bad in my
opinion!

Aside from the fact it's not pretty, there is a reason for having nice
separate layers, the fact that you can re-use / re-write the 'old buffer'
and do need to use unnecessary extra space. Violating that contract by
letting layers access the full data of another layer it should not see,
requires extra storage space, which on most embedded systems is very scarce.

I understand the issue, but in my opinion this is not the way to solve it.
So please reconsider solving the issue this way.

Paul