Re: [lamps] Call for adoption for draft-ito-documentsigning-eku

["Blumenthal, Uri - 0553 - MITLL" <[e-mail]>]

On Thu, Jul 29, 2021 at 11:54 AM Blumenthal, Uri - 0553 - MITLL <[e-mail]> wrote:

“A bazillion of EKUs” is a much greater risk, in my opinion. Which I’d like to avoid at all costs.

 

> Since the scenario here is largely hyperbolic, entirely unrealistic, 

 

I think it’s very realistic, and only mildly hyperbolic. What I do think is hyperbolic and not entirely realistic is your concern of cross-format threats.

 

>  .  .  .  and I think reflects a misunderstanding of the proposal, perhaps you can clarify:

> Do you share the same concerns regarding certificatePolicies? 

 

As a matter of fact, I do not [share the same concerns], and here’s why:

 
KU usually is marked Critical, and everybody is expected to correctly grok it or die;
EKU is sometimes marked Critical, and everybody should grok it correctly – there’s reasonably small set of them, covering reasonable expected use cases;
Certificate policies are rarely marked Critical, they cover <whatever>, very often admin domain-specific, and if an app cannot grok it – the common practice is “just ignore and proceed”.
 

I would hate to see EKU proliferating in quantity and “degrading” in quality to the point of “I don’t really care …”

 

For me and my use cases it is very convenient to distinguish a cert for authenticating from a cert for signing documents and/or emails, but absolutely counter-productive to go OCD-route and distinguishing between signing PDF, signing Word, signing XML, signing ROT-13… Plus, as I said earlier – my use case involves hardware tokens that have only one certificate slot for Digital Signature. In theory I could load that one cert with 2KB of EKUs, but it makes zero sense.

 

It looks to me like there are organizations that have very narrow use cases, e.g., just signing PDF – and they’d like to be able to set one EKU on a single cert that they want to issue, to do exactly that and nothing more. I can understand that.

There are other organizations whose use cases are broader than the above. Their options are: (a) EKUs that are not “too narrow”, (b) cram signing cert with a ton of EKUs to address all the use cases, and (c) issue a ton of certs, each for specific use case. It seems that the preference of the majority is (a), and for a good practical reason. As I explained earlier, cert management is costly.