Looking at it another way, the timelines are:

 

24 hours if the Subscriber requests the cert (no certificate problem report)

48 hours if there is a key compromise (24 hour investigation + 24 hour to 
revoke)

8 days if the cert was issued to the wrong domain name or organization (7 day 
investigation + 24 hours to revoke) *

14 days for all other reasons

 

* My heartburn over how long this is to take care of.  8 days is a long time 
where domain validation failed.

 

I think the requirement to reply to the certificate problem report is built in 
by requiring the CA to work with the entity making the report. I don’t have a 
good idea on how to improve the escalation path.

 

From: Ryan Sleevi [mailto:sle...@google.com] 
Sent: Wednesday, August 23, 2017 1:33 PM
To: Jeremy Rowley <jeremy.row...@digicert.com>
Cc: CA/Browser Forum Public Discussion List <public@cabforum.org>
Subject: Re: [cabfpub] Revocation ballot v2

 

 

 

On Wed, Aug 23, 2017 at 3:25 PM, Jeremy Rowley <jeremy.row...@digicert.com 
<mailto:jeremy.row...@digicert.com> > wrote:

Hmm  - that does seem long.  What if we keep the investigation to 24 hours and 
change revocation to 24 hours/2 weeks? There’s no reason for the CA to delay 
investigating any issue.

 

I wasn't trying to suggest it's long. I think we've seen some CAs want to 
investigate, such as relevant RFCs, errata, document and change control 
history, etc. That it is bounded at an upper-bound (by 14 days) keeps it within 
a reasonable frame. If a CA wants more time, they should be proactively 
internally reviewing for compliance :)

 

For transparency, what do you suggest?  I left it the same as today. Perhaps 
state that the CA MUST reply to the certificate problem reporter about its 
decision within 3 days?

 

I know we'd previously talked about reports to public@cabforum.org 
<mailto:public@cabforum.org> , in line with how 9.16.3 is handled for local 
law, but the fact that we have a maximum cap at 14 days for revocation does, I 
think, reduce the risk. I'm also sensitive to the fact that running any form of 
'security bug queue' will result in absolutely terrible submissions that are 
fundamentally incorrect, and I don't want to further impose additional costs 
for having to deal with junk-reports.

 

My concern is that CAs who receive problem reports would be incentivized to 
close them as "not an issue", and thus force the root stores to get involved, 
but that's arguably the status quo today. I think root stores that want to 
address this would have a variety of tools - for example, requiring quarterly 
disclosures of the number of problem reports received, responded to, etc - that 
could address some of those high-level concerns, hopefully without introducing 
significant burden for junk reports. After all, today, if a CA gets a junk 
report, there's also zero disclosure requirements - so I'm not terribly 
convinced we need to solve that problem as well, provided that the cap at 14 
days stands.

 

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
Public mailing list
Public@cabforum.org
https://cabforum.org/mailman/listinfo/public

Reply via email to