Re: [secdir] Secdir review of draft-ietf-tls-record-limit

[Benjamin Kaduk <[e-mail]>]

On Fri, Feb 23, 2018 at 09:49:11AM -0500, Alan DeKok wrote:
> On Feb 22, 2018, at 8:19 PM, Martin Thomson <[e-mail]> wrote:
> > A few changes based on your feedback here:
> > https://github.com/tlswg/tls-record-limit/pull/17
> 
>   Thanks. That helps.
> 
> > The remainder of the paragraph is intended to make that distinction
> > clear.  i.e., this limit is inherently fixed by negotiation, the other
> > is dynamic.
> 
>   My point is that those two limits may agree or disagree dynamically.  So it would be good to give guidance on what to do when a previous agreement dynamically changes to disagreement.
> 
> >> Comment:  it would be good to give guidance on what to do here, and what happens in error cases.
> > 
> > DTLS (RFC 6347) already has some fairly extensive guidance on PMTUD,
> > which I didn't want to replicate here.  You hit a lot of the issues in
> > your questions.  With some better citations, this is what I came up
> > with:
> 
>   That still doesn't give guidance.  What happens if the record size limit is fine at the start of a session, and then MTU changes, and the packets no longer make it through?   What does the application do?
> 
>   RFC 6347 Section 4.1.1.1 gives some guidance, but I think not enough.  Exposing the MTU to the application is good, but what does the application *do* with this information?
> 
>   e.g. PMTU should be exposed to the application as per Section 4.1.1.1 of {{?DTLS}}.  If the PTMU changes, the application may discover that the new MTU is smaller than the record size limit.  In that situation, the only recourse available may be to close the session, and to open a newer one with a smaller record size limit that is compatible with the new MTU.

The record size limit is a cap, not a mandatory record size.  So,
the application could just send smaller records that fit within the
new PMTU, provided both sides have the proper logic around PMTU
discovery and record sizing.

-Benjamin