On Thu, Jul 29, 2021 at 5:15 PM Salz, Rich <[email protected]> wrote:

>
>    - I'm not sure this is correct, Rich? Logs regularly rotate IDs;
>    presently annually, but it's reasonable to anticipate more frequently as
>    the size/performance tradeoffs, precisely as the way of pruning the 
> storage.
>
>
>
> Thanks for the correction!
>
>
>
> So how do replying parties know the log ID changed?
>

To be clear, the log ID doesn't change (and refer to the same log). Each
Log ID uniquely identifies a log. For example, if a log changes a key, it's
functionally a new log - this is true in 6962 as it is in 6962-bis. The
only thing that changed in -bis is from identifying logs by key hashes to
identifying by OIDs, which was meant to make for smaller encoding.

The draft flags that the communication of LogIDs is fundamentally something
that is done out-of-band -
https://www.ietf.org/archive/id/draft-ietf-trans-rfc6962-bis-40.html#name-log-id
- i.e. up to client policy. For example, for two widespread 6962
implementations (Apple's {mac, i, tv}OS and Google's Chrome), the list of
recognized logs is governed by user agent/vendor policy, and those logs are
communicated by the vendor (aka "Trusted Log Lists", although trust is a
bit of a stretch)


> And do we need to give Log’s not an ID but rather an arc?
>

No. As with 6962, the idea here is not that "This log was ID X, and is now
ID Y" - but rather "I know of a log with ID X, and I now know of a log with
ID Y, and these are logically distinct, even if their contents are
identical"
_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans

Reply via email to