Re: [saag] [tsvwg] Comments on draft-ietf-tsvwg-transport-encrypt-08.txt

[Peter Gutmann <[e-mail]>]

Tom Herbert <[e-mail]> writes:

>https://www.iab.org/wp-content/IAB-uploads/2014/12/semi2015_huitema.pdf
>describes with some detail how encryption of the transport layer is
>beneficial to resolve the tussel that results in protocol ossification.

I'd always assumed the push for header encryption was due to some vague
handwaving about privacy or something else that's never really been made
clear, but from the above document it looks like the motivation is purely
ideological ("If we want to break that ossification, we have to do something
drastic", "we can score at least a partial win for end-to-end connectivity",
"application developers win a similar tussle", "reverses the balance of power
between end-to-end and middle-boxes"), which indicates that it'll end up as a
repeat of the IPsec debacle ("Let's make sure IPsec breaks NAT, because IPsec
will be bigger than NAT and so NAT will have to go away").

Someone really needs to do a proper design document and threat model for this
if it's that unclear what the goals of header protection are - is it for some
sort of privacy thing, an ideological move because people don't like
middleboxes, something to thwart the lizard people, to impress Jodie Foster,
or what?

Might I suggest that the IETF suspend any further push towards header
encryption until someone can figure out:

a. Why we're doing this.

b. What its intended goals and effects are.

c. Whether those goals will be met in practice.

d. Whether it will work when deployed in the real world.

At the moment it looks like (a) is at best very unclear and (b), (c), and (d)
are pretty much nonexistent.

Peter.