Re: [TLS] Encrypt content type (#51) + Remove TLSCiphertext.version (#144)

[Eric Rescorla <[e-mail]>]

The issue is that there are believed to be middleboxes which depend on the
length being
in the right place.

-Ekr


On Sat, May 16, 2015 at 6:04 PM, Dave Garrett <[e-mail]>
wrote:

> On Saturday, May 16, 2015 08:23:19 pm Eric Rescorla wrote:
> > On Sat, May 16, 2015 at 4:54 PM, Dave Garrett <[e-mail]>
> > wrote:
> > > Is there enough agreement on list for consensus on:
> > > https://github.com/tlswg/tls13-spec/pull/51/files
> > > &
> > > https://github.com/tlswg/tls13-spec/issues/144
> > > ?
> > >
> > > This would get us to the point of no AD for AEAD, as noted by Martin
> > > Thomson.
> > > https://www.ietf.org/mail-archive/web/tls/current/msg15490.html
> >
> > No. I believe that the present consensus is to retain the existing
> version
> > bytes in place but just leave them fixed.
>
> To be clear, the proposal would be to drop TLSCiphertext.version, but NOT
> TLSPlaintext.version. The latter needs to stay in place for compatibility,
> but the former is after version negotiation and provides no useful info.
> Just a
> waste of 2 bytes for every record at that point. The version number is
> currently in the context string for the certificate verification
> signatures, and
> Martin suggested mixing it into the PRF/HKDF info string to make the keying
> material dependent on the version number. (see quoted mailing list URL)
>
>
> Dave
>