Re: [TLS] rfc7366: is encrypt-then-mac implemented?

[Peter Gutmann <[e-mail]>]

Manuel Pégourié-Gonnard <[e-mail]> writes:

>Another (related) idea: explicitly change the definition of
>GenericBlockCipher when EtM is in use

Henrick had suggested something like that as well, but at that point you're
starting to rewrite chunks of the main TLS RFCs, which may be overkill if the
goal is merely to explain which length value to use.  That is, it's valuable
for a future RFC (e.g. TLS 1.3) where you'd explicitly want to make this the
standard record format, but it may be a bit heavy going for an erratum that's
only supposed to say "use this value for the length".

The ideal explanation would be as short and concise as possible, and one where
no-one has any concerns about different interpretations.  This was why I was
leaning towards the "for the avoidance of any doubt ..." style of annotation,
it's a brief clarification to the existing text to clear up ambiguity.  My
concern is that adding a complex technical definition, while very precise, may
then require another erratum later on to explain it.

If anyone can come up with a clear one-paragraph note (which, when posted to
the list, no-one complains about :-), I'd be happy to use that.

Peter.