Re: [Trans] Relaxing section 5.1

[Peter Bowen <[e-mail]>]

On Wed, Nov 2, 2016 at 6:39 AM, Melinda Shore <[e-mail]> wrote:
> 6962-bis has completed working group last call.  Minor editorial
> changes are fine, but let's try to avoid major changes that would
> require yet another WGLC.  If there's a need for an additional
> document dealing with operational considerations or operational
> specifications, we can do that.  If there's a major problem with
> 6962-bis, we can deal with that as well, but it would need to
> be serious (i.e. goes to the correctness of the specification,
> fixes something that's broken, etc.).

I realize that 6962bis has passed WGLC, so I know there is a high bar
for changes.  However I think this might pass that bar.  The highly
restrictive language that imposes minimum policy for logs prevents
interoperability with other IETF RFCs on the standards track very
hard.  6962bis appears to assume that DANE (RFCs 7671 and 6698) will
never be implemented and that concepts like RFC 6091 will never come
to fruition.

By requiring all logs MUST accept any certificate that chains to a
root in the log's root list, 6962bis fails to allow log operators to
mitigate any Denial of Service attacks mounted by attempting to log
massive numbers of certificates that are not relevant to the log
scope.  For example, many existing certification authorities issue
both server authentication certificates and certificates for personal
identification.  For some roots, acquiring large numbers of these is
relatively easy (see discussion of fetching millions of Taiwanese
Citizen Digital Certificates in
https://smartfacts.cr.yp.to/smartfacts-20130916.pdf)  As written
today, a log MUST accept these.  There is no option for a log to
require that all certificates must meet some usability criteria.

I agree that a future additional document dealing with operational
considerations is fine, but as drafted today, 6962bis does not allow a
log to implement these considerations.

Thanks,
Peter