search

dennisjackson / trust-negotiation-comments

2 stars 0 forks source link

Incorrect analysis of rotating root keys in usecases doc #17

Closed nharper closed 3 months ago

nharper commented 3 months ago

The usecases document states that its purpose is the following:

The analysis looks to evaluate:

  • Is the proposed problem compelling? Does it need to be solved?
  • To what extent do the drafts solve the problem and under what deployment assumptions?
  • How does it compare to the alternatives?

My reading of the "Rotating Root Keys" section is that you're saying the problem is not compelling, and root keys don't need to be rotated. Is that correct?

The drafts present rotating root keys as a problem. Even if you think the problem isn't compelling, by changing the framing of your analysis to a different problem, you are failing to evaluate how the drafts solve the stated problem of root key rotation. If the usecases document is going to serve its intended purpose, it needs to evaluate the use cases presented in the drafts.

(This is a follow-up on issue #5. See that issue for details on how the "Rotating Root Keys" section ignores the problem that root key rotation is something that happens and will continue to happen, and how the section instead proposes a different solution that does not rotate root keys.)

dennisjackson commented 3 months ago

The analysis already answers all three questions regarding key rotation:

Is the proposed problem compelling?

No. It's a symptom of a different problem (security risks of long-lived keys).

To what extent do the drafts solve the problem and under what deployment assumptions?

It functions adequately for clients and server who adopt T.E. / T.A.

How does it compare to the alternatives?

Poorly. Cross-Signings works just as well and is already ubiquitously supported.

These answers are already reflected in the text.

nharper commented 3 months ago

Is the proposed problem compelling?

No. It's a symptom of a different problem (security risks of long-lived keys).

Understood. Despite it going against security best practices, existing Mozilla root program policy, and Chrome root program vision, you're stating that the Web PKI shouldn't rotate root keys.

dennisjackson commented 3 months ago

The literal text of the document you're commenting on:

root key rotation is fairly unimportant for the security of clients (whereas intermediate certificate key rotation is vital)

Please take some time to cool down.