[TLS] Re: Discussion about the Deployment Decision

Bas Westerbaan <[email protected]> Mon, 3 Aug 2026 16:44:57 +0200
Newsgroups gmane.ietf.tls
Message-ID <CAMjbhoUeGzikXOKZoD_=F0XSLkJdpEHL8UK=e99GY5FKUuNLiA@mail.gmail.com>
--===============7906285613629020341==
Content-Type: multipart/alternative; boundary="000000000000afc3590658259534"

--000000000000afc3590658259534
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Mon, Aug 3, 2026 at 1:02=E2=80=AFPM Paul Romer <romerp=3D40bc.edu@dmarc.=
ietf.org>
wrote:

> Sophie, you wrote:
> > A recommendation flag should do nicely to convey expected utility.
>
> You also wrote:
> > Well, technically my opinion is that X25519MLKEM768 should be
> recommended,
> > but I also think that neither this flag nor the MTI flag have any meani=
ng
> > in the first place
>
> https://mailarchive.ietf.org/arch/msg/tls/anqTfO9uoZXfHTOeiV1wZOm1GVo/
>
> Are you trolling me?
>
> I don't see how this group will be able to establish a consensus about
> anything if everyone just looks at their shoes when you claim that
>
> - A is true because P =3D> A
>
> and
>
> - B is true because =C2=ACP =3D> B
>
> where
>
> P  =3D "The recommended flag conveys information"
> =C2=ACP =3D "The recommended flag has no meaning"
>
>
> Look, anyone can make a mistake. It's no big deal as long as the rest
> of the group does it's job:
>
> - flag any logical inconsistencies and any misleading or incorrect
> statements of fact;
>
> - push for logical coherence and alignment with the evidence.
>
> The deep problem is the missing messages. People seem to be afraid to
> speak up about the mistakes.
>

Calling out a Nobel laureate indeed requires some nerve. I think it's
pretty obvious that two seemingly contradictory statements can make total
sense when considered in context.



>
> On Fri, Jul 31, 2026 at 1:42=E2=80=AFPM Sophie Schmieg
> <[email protected]> wrote:
> >
> > I think it's overall probably prudent to not assume implementers have a
> degree or even interest in economy. A recommendation flag should do nicel=
y
> to convey expected utility.
> >
> > On Fri, Jul 31, 2026 at 10:24=E2=80=AFAM Paul Romer <romerp=3D
> [email protected]> wrote:
> >>
> >> log( ) is concave. I should have written "concave utility function"
> >> instead of "convex utility function." I tripped over the change of
> >> sign associated with the use of a utility function instead of a loss
> >> function.
> >>
> >> On Fri, Jul 31, 2026 at 12:53=E2=80=AFPM Paul Romer <[email protected]> wr=
ote:
> >> >
> >> > The response to my message with the subject "Improving the Quality
> >> > ..." has split into two different discussions:
> >> >
> >> > 1. One addresses the decision to deploy MLKEM768 instead of
> X25519MLKEM768.
> >> >
> >> > 2. The second is concerned with the tally of support for publishing
> the RFC
> >> >
> >> > I'm starting a new thread here that can focus on discussion 1 to
> >> > highlight some positive contributions to that discussion.
> >> >
> >> >
> >> > - Mark pointed out that I made a mistake by referring to a "Feynman
> >> > estimate". I should have said a "Fermi estimate." He is right and hi=
s
> >> > message shows why a focused response about a specific issue can be s=
o
> >> > useful in discussions like this. I made a mistake. If no one had
> >> > corrected it, it could have encouraged others to repeat that mistake=
.
> >> > Mark corrected the mistake. We now have a consensus about how to ref=
er
> >> > to this type of estimation. It is easier to build consensus with a
> >> > series of small steps like this one that focus on narrow specifics.
> >> >
> >> >
> https://mailarchive.ietf.org/arch/msg/tls/qAPFBiBwPPXj0nST62FfJ3PsDCU/
> >> >
> >> >
> >> > - Dennis said that I repeated arguments from my original post and th=
at
> >> > it is important to avoid repetition. On reflection, he is correct on
> >> > both counts and I will try not to make this mistake again.
> >> >
> >> >
> https://mailarchive.ietf.org/arch/msg/tls/vw685VPl4aXLWmqgOfdEzyWyADw/
> >> >
> >> >
> >> > - Dennis also pointed to an alternative way to estimate the benefit =
of
> >> > deploying MLKEM768. What would be helpful would be some rough
> >> > calculations that show how the extension he has in mind affects the
> >> > comparison of costs and benefits.
> >> >
> >> >
> >> > - Dennis objected to the admittedly ad hoc method that I used to com=
e
> >> > up with an estimate of the cost of the reduction in security provide=
d
> >> > by MLKEM768 instead of X25519MLKEM768. In light of the news about th=
e
> >> > Anthropic attack on HAWK, an interesting way to examine the costs of
> >> > changes in security would be to allow for a convex utility function
> >> > that captures the idea of risk aversion. Instead of using the implic=
it
> >> > decision rule "make the change if B > C", a better rule would be mak=
e
> >> > the change if
> >> >
> >> >          E(U(B-C)) > 0,
> >> >
> >> > where E is the expectations operator and U is the utility function. =
A
> >> > natural initial choice is logarithmic utility U(y) =3D ln(y).
> >> >
> >> > For this type of estimation, you have to have a stomach for extreme
> >> > simplification. One might start by continuing to treat B as
> >> > deterministic and allowing for just two states of the world with
> >> > different costs C_1 and C_2. With probability p there is no successf=
ul
> >> > attack on MLKEM768 and the ex post cost of deploying it is C_1. With
> >> > probability (1-p) there is a successful attack and the ex post cost =
is
> >> > C_2 > C_1.
> >> >
> >> > This does not address an issue that Donald Bernstein has emphasized,
> >> > that there is also a risk of implementation errors in software that
> >> > supports MLKEM768. That could be added into a subsequent analysis.
> >> > Given the news about the attack on Hawk, it seems reasonable to focu=
s
> >> > first on an analysis of the effects of a change in the probability o=
f
> >> > a successful attack on the protocol.
> >> >
> >> > In the simple framework that I suggest, one can do a sensitivity
> >> > analysis that varies the estimates of C_1 and C_2, but the more
> >> > interesting analysis would fix those values and explore small change=
s
> >> > in p.
> >> >
> >> > I think it might not be helpful for me to contribute an analysis alo=
ng
> >> > these lines. It is obvious that I think it would be a mistake to
> >> > deploy MLKEM768. Allowing for risk aversion will make a decision to
> >> > deploy MLKEM768 look worse. A neutral observer might worry that any
> >> > analysis I do is biased by motivated reasoning. It might be helpful
> >> > for someone else to give this a try.
> >>
> >> _______________________________________________
> >> TLS mailing list -- [email protected]
> >> To unsubscribe send an email to [email protected]
> >
> >
> >
> > --
> >
> > Sophie Schmieg | Information Security Engineer | ISE Crypto |
> [email protected]
> >
>
> _______________________________________________
> TLS mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
>

--000000000000afc3590658259534
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote g=
mail_quote_container"><div dir=3D"ltr" class=3D"gmail_attr">On Mon, Aug 3, =
2026 at 1:02=E2=80=AFPM Paul Romer &lt;romerp=3D<a href=3D"mailto:40bc.edu@=
dmarc.ietf.org">[email protected]</a>&gt; wrote:<br></div><blockquote=
 class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px so=
lid rgb(204,204,204);padding-left:1ex">Sophie, you wrote:<br>
&gt; A recommendation flag should do nicely to convey expected utility.<br>
<br>
You also wrote:<br>
&gt; Well, technically my opinion is that X25519MLKEM768 should be recommen=
ded,<br>
&gt; but I also think that neither this flag nor the MTI flag have any mean=
ing<br>
&gt; in the first place<br>
<br>
<a href=3D"https://mailarchive.ietf.org/arch/msg/tls/anqTfO9uoZXfHTOeiV1wZO=
m1GVo/" rel=3D"noreferrer" target=3D"_blank">https://mailarchive.ietf.org/a=
rch/msg/tls/anqTfO9uoZXfHTOeiV1wZOm1GVo/</a><br>
<br>
Are you trolling me?<br>
<br>
I don&#39;t see how this group will be able to establish a consensus about<=
br>
anything if everyone just looks at their shoes when you claim that<br>
<br>
- A is true because P =3D&gt; A<br>
<br>
and<br>
<br>
- B is true because =C2=ACP =3D&gt; B<br>
<br>
where<br>
<br>
P=C2=A0 =3D &quot;The recommended flag conveys information&quot;<br>
=C2=ACP =3D &quot;The recommended flag has no meaning&quot;<br>
<br>
<br>
Look, anyone can make a mistake. It&#39;s no big deal as long as the rest<b=
r>
of the group does it&#39;s job:<br>
<br>
- flag any logical inconsistencies and any misleading or incorrect<br>
statements of fact;<br>
<br>
- push for logical coherence and alignment with the evidence.<br>
<br>
The deep problem is the missing messages. People seem to be afraid to<br>
speak up about the mistakes.<br></blockquote><div><br></div><div>Calling ou=
t a Nobel laureate indeed requires some nerve. I think it&#39;s pretty obvi=
ous that two seemingly contradictory statements can make total sense when c=
onsidered in context.</div><div><br></div><div>=C2=A0</div><blockquote clas=
s=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid r=
gb(204,204,204);padding-left:1ex">
<br>
On Fri, Jul 31, 2026 at 1:42=E2=80=AFPM Sophie Schmieg<br>
&lt;sschmieg=3D<a href=3D"mailto:[email protected]" target=3D"_bl=
ank">[email protected]</a>&gt; wrote:<br>
&gt;<br>
&gt; I think it&#39;s overall probably prudent to not assume implementers h=
ave a degree or even interest in economy. A recommendation flag should do n=
icely to convey expected utility.<br>
&gt;<br>
&gt; On Fri, Jul 31, 2026 at 10:24=E2=80=AFAM Paul Romer &lt;romerp=3D<a hr=
ef=3D"mailto:[email protected]" target=3D"_blank">[email protected]=
.org</a>&gt; wrote:<br>
&gt;&gt;<br>
&gt;&gt; log( ) is concave. I should have written &quot;concave utility fun=
ction&quot;<br>
&gt;&gt; instead of &quot;convex utility function.&quot; I tripped over the=
 change of<br>
&gt;&gt; sign associated with the use of a utility function instead of a lo=
ss<br>
&gt;&gt; function.<br>
&gt;&gt;<br>
&gt;&gt; On Fri, Jul 31, 2026 at 12:53=E2=80=AFPM Paul Romer &lt;<a href=3D=
"mailto:[email protected]" target=3D"_blank">[email protected]</a>&gt; wrote:<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; The response to my message with the subject &quot;Improving t=
he Quality<br>
&gt;&gt; &gt; ...&quot; has split into two different discussions:<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; 1. One addresses the decision to deploy MLKEM768 instead of X=
25519MLKEM768.<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; 2. The second is concerned with the tally of support for publ=
ishing the RFC<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; I&#39;m starting a new thread here that can focus on discussi=
on 1 to<br>
&gt;&gt; &gt; highlight some positive contributions to that discussion.<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; - Mark pointed out that I made a mistake by referring to a &q=
uot;Feynman<br>
&gt;&gt; &gt; estimate&quot;. I should have said a &quot;Fermi estimate.&qu=
ot; He is right and his<br>
&gt;&gt; &gt; message shows why a focused response about a specific issue c=
an be so<br>
&gt;&gt; &gt; useful in discussions like this. I made a mistake. If no one =
had<br>
&gt;&gt; &gt; corrected it, it could have encouraged others to repeat that =
mistake.<br>
&gt;&gt; &gt; Mark corrected the mistake. We now have a consensus about how=
 to refer<br>
&gt;&gt; &gt; to this type of estimation. It is easier to build consensus w=
ith a<br>
&gt;&gt; &gt; series of small steps like this one that focus on narrow spec=
ifics.<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; <a href=3D"https://mailarchive.ietf.org/arch/msg/tls/qAPFBiBw=
PPXj0nST62FfJ3PsDCU/" rel=3D"noreferrer" target=3D"_blank">https://mailarch=
ive.ietf.org/arch/msg/tls/qAPFBiBwPPXj0nST62FfJ3PsDCU/</a><br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; - Dennis said that I repeated arguments from my original post=
 and that<br>
&gt;&gt; &gt; it is important to avoid repetition. On reflection, he is cor=
rect on<br>
&gt;&gt; &gt; both counts and I will try not to make this mistake again.<br=
>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; <a href=3D"https://mailarchive.ietf.org/arch/msg/tls/vw685VPl=
4aXLWmqgOfdEzyWyADw/" rel=3D"noreferrer" target=3D"_blank">https://mailarch=
ive.ietf.org/arch/msg/tls/vw685VPl4aXLWmqgOfdEzyWyADw/</a><br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; - Dennis also pointed to an alternative way to estimate the b=
enefit of<br>
&gt;&gt; &gt; deploying MLKEM768. What would be helpful would be some rough=
<br>
&gt;&gt; &gt; calculations that show how the extension he has in mind affec=
ts the<br>
&gt;&gt; &gt; comparison of costs and benefits.<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; - Dennis objected to the admittedly ad hoc method that I used=
 to come<br>
&gt;&gt; &gt; up with an estimate of the cost of the reduction in security =
provided<br>
&gt;&gt; &gt; by MLKEM768 instead of X25519MLKEM768. In light of the news a=
bout the<br>
&gt;&gt; &gt; Anthropic attack on HAWK, an interesting way to examine the c=
osts of<br>
&gt;&gt; &gt; changes in security would be to allow for a convex utility fu=
nction<br>
&gt;&gt; &gt; that captures the idea of risk aversion. Instead of using the=
 implicit<br>
&gt;&gt; &gt; decision rule &quot;make the change if B &gt; C&quot;, a bett=
er rule would be make<br>
&gt;&gt; &gt; the change if<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt;=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 E(U(B-C)) &gt; 0,<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; where E is the expectations operator and U is the utility fun=
ction. A<br>
&gt;&gt; &gt; natural initial choice is logarithmic utility U(y) =3D ln(y).=
<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; For this type of estimation, you have to have a stomach for e=
xtreme<br>
&gt;&gt; &gt; simplification. One might start by continuing to treat B as<b=
r>
&gt;&gt; &gt; deterministic and allowing for just two states of the world w=
ith<br>
&gt;&gt; &gt; different costs C_1 and C_2. With probability p there is no s=
uccessful<br>
&gt;&gt; &gt; attack on MLKEM768 and the ex post cost of deploying it is C_=
1. With<br>
&gt;&gt; &gt; probability (1-p) there is a successful attack and the ex pos=
t cost is<br>
&gt;&gt; &gt; C_2 &gt; C_1.<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; This does not address an issue that Donald Bernstein has emph=
asized,<br>
&gt;&gt; &gt; that there is also a risk of implementation errors in softwar=
e that<br>
&gt;&gt; &gt; supports MLKEM768. That could be added into a subsequent anal=
ysis.<br>
&gt;&gt; &gt; Given the news about the attack on Hawk, it seems reasonable =
to focus<br>
&gt;&gt; &gt; first on an analysis of the effects of a change in the probab=
ility of<br>
&gt;&gt; &gt; a successful attack on the protocol.<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; In the simple framework that I suggest, one can do a sensitiv=
ity<br>
&gt;&gt; &gt; analysis that varies the estimates of C_1 and C_2, but the mo=
re<br>
&gt;&gt; &gt; interesting analysis would fix those values and explore small=
 changes<br>
&gt;&gt; &gt; in p.<br>
&gt;&gt; &gt;<br>
&gt;&gt; &gt; I think it might not be helpful for me to contribute an analy=
sis along<br>
&gt;&gt; &gt; these lines. It is obvious that I think it would be a mistake=
 to<br>
&gt;&gt; &gt; deploy MLKEM768. Allowing for risk aversion will make a decis=
ion to<br>
&gt;&gt; &gt; deploy MLKEM768 look worse. A neutral observer might worry th=
at any<br>
&gt;&gt; &gt; analysis I do is biased by motivated reasoning. It might be h=
elpful<br>
&gt;&gt; &gt; for someone else to give this a try.<br>
&gt;&gt;<br>
&gt;&gt; _______________________________________________<br>
&gt;&gt; TLS mailing list -- <a href=3D"mailto:[email protected]" target=3D"_bla=
nk">[email protected]</a><br>
&gt;&gt; To unsubscribe send an email to <a href=3D"mailto:[email protected]=
rg" target=3D"_blank">[email protected]</a><br>
&gt;<br>
&gt;<br>
&gt;<br>
&gt; --<br>
&gt;<br>
&gt; Sophie Schmieg | Information Security Engineer | ISE Crypto | <a href=
=3D"mailto:[email protected]" target=3D"_blank">[email protected]</a><b=
r>
&gt;<br>
<br>
_______________________________________________<br>
TLS mailing list -- <a href=3D"mailto:[email protected]" target=3D"_blank">tls@i=
etf.org</a><br>
To unsubscribe send an email to <a href=3D"mailto:[email protected]" targe=
t=3D"_blank">[email protected]</a><br>
</blockquote></div></div>

--000000000000afc3590658259534--


--===============7906285613629020341==
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: base64
Content-Disposition: inline

X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KVExTIG1haWxp
bmcgbGlzdCAtLSB0bHNAaWV0Zi5vcmcKVG8gdW5zdWJzY3JpYmUgc2VuZCBhbiBlbWFpbCB0byB0
bHMtbGVhdmVAaWV0Zi5vcmcK

--===============7906285613629020341==--