Rob,

Your reply did not contain any info that would cause me to change my mind
about the staging period. I think we agree that the staging period cannot
guarantee that all RPs have the new CA cert before the old data is replaced.
What it can do is increase the likelihood that this will happen. Your position seems to be that the staging period may cause more harm than good. I appreciate that notion, but disagree.

I think that changing the MUST to a SHOULD (or to RECOMMEND), and noting that an emergency rekey is not bound by this rule is a reasonable compromise. I am not willing to go with MAY; it may encourage bad behavior.

BBN does not produce CA/repository management code, so we are not arguing from a position of not wanting to change our code. We are arguing from a position of trying to minimize the likelihood of RP confusion.

Your description of how your CA code works omits a few details. For example, in step 1, when you start issuing new signed products, these replace the old products, because of the naming conventions employed in the repository. Thus you seem to be creating a mixed (inconsistent) set of repository entries during this step. The staging period, as described in the keyroll doc, provides more detail about this process, and shows how to minimize the window of inconsistency. That strikes me as a fundamentally good thing to do.

I acknowledge that offering an option to restart the rollover during the
staging period add complexity. But, there is also added complexity for RPs who have to deal with repository fetches that yield a mix of old and new signed products. Saying that an RP will have to deal with this issue in some cases does not mean that we ought not try hard to minimize the frequency with which RPs have to do so. This seems especially true since we seem to lack a good description of how an RP should deal with this problem. I have seen no algorithmic description of how an RP is supposed to deal with repository inconsistency. Not even heuristics.)

Steve

P.S.  BTW, no, our RP code does not discard old data before getting new data.
Has your validator stopped beating its database :-)?
_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr

Reply via email to