[Anima] Re: [Uta] Re: [Last-Call] Re: Concern about draft-ietf-uta-require-tls13-10 with IoT protocols

[Eric Rescorla <[e-mail]>]

On Thu, Apr 10, 2025 at 6:54 AM Michael Richardson <[e-mail]> wrote:

> Peter Gutmann <[e-mail]> wrote:
>     > Some sort of qualification like that would be my preference as
> well.  I don't
>     > think I've ever encountered TLS 1.3 in SCADA (I mean, there's still
> a lot of
>     > TLS 1.0 out there that people are struggling to move to TLS 1.2), so
> you could
>     > just as easily say "TLS 1.3 has next to no deployment" depending on
> your terms
>     > of reference.
>
> Yes.
> TLS 1.3 is ubiquitous, so why is TLS 1.2 still enabled at all major sites?
>

First, the document doesn't say ubiquitous. It says "in widespread use".

Second, in response to the question "why do sites still support TLS 1.2".
Web browser vendors and Web sites are very conservative about breakage,
and even small breakage fractions are sufficient to not make a change.
As an example, when TLS 1.3 was being designed, measurements showed
that we would see an increase in connection failures of a few percentage
points, which motivated a fair amount of last minute design work to bring
the failure rate into line with TLS 1.2.

As long as there is any significant fraction (by which I mean in excess
of fractions of a percent) of the client population that
supports TLS 1.2 only, I would expect servers to continue to support
it. That's entirely consistent with TLS 1.3 being in widespread use.

-Ekr