On Mar 28, 2022, at 8:56 AM, Michael Richardson <[email protected]> wrote:
> I understand.
> For mechanisms like BRSKI (and CSP/MATTER, and probably UPC UA) where the end
> product of onboarding is a certificate issued to the device, then the
> question about when to refresh is mostly given by the notAfter parameter.
> In most cases one should start refreshing those credentials at the 50% mark.
> 
> The ACME WG is discussing having the certificate renew point be better
> specified (in the ACME protocol), but I think that we will learn from this
> and wind up with something in the certificate if the protocol can't carry
> additional meta-data.

  That's a good chunk of what I'm proposing in my document.  There are often 
existing cert fields which contain the information needed for configuration.  
It would be nice to use them.

> There are advantages of having renewals spread across time, but there are
> also disadvantages as it spreads the failure signal across time as well which
> makes it harder to see that there is a real problem.

  Generally if people can't renew, the server should log errors, or the end 
user device has a real network where it can report errors.

  Alan DeKok.

_______________________________________________
Emu mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/emu

Reply via email to