-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 09/18/2012 10:08 AM, André Becker wrote:
> Hi Marc,
> 
> thanks for the quick explanation, I'm getting your point. Even without the
> OAM use case in mind I would support the idea of using an EKU to mark
> RELOAD certificates as what they are. From my point of view this definitley
> belongs into the base draft. And, as the EKU's purpose in the OAM use case
> isn't just to allow the RELOAD-usage, but rather to forbid others, the
> cleanest approach should be to make the usage of the EKU mandatory and to
> mark it as critical. The additional implementation effort seems acceptable
> to me.

OK, that's a different opinion than Thomas', so we need more discussion on the
specific point of having it in -base or as an extension.  Because I want to
make it as painless as possible to the -base authors, I'll still publish my
draft as an extension, so the -base author can just copy and paste the text if
we all agree to put this in -base.

As for my opinion on this subject, I agree that -base is the right place to
put this, but seeing that the things we agreed on nearly 2 months ago in
Vancouver still did not make it in the spec, I fear that the only path to
success is an extension.

> 
> Regards, André
> 
> Am 18.09.2012 17:53, schrieb Marc Petit-Huguenin: Hi André,
> 
> On 09/18/2012 08:37 AM, André Becker wrote:
>>>> Hi,
>>>> 
>>>> I think there's a small point I don't get. You write that in the
>>>> current situation the server would need two different self-signed
>>>> certificates for the use case: One for signing reload certificates,
>>>> one for signing https client certificates.
>>>> 
>>>> However, in my understanding, if you self-sign a certificate, there's
>>>> no one who tells you what to do or not to do with it. Especially, you
>>>> may sign whatever you want. So why do you need two self-signed
>>>> certificates after all?
> Let's say that I use only one CA to sign both certificates for RELOAD and
> for the Web API.  In this case, a certificate signed for the Web API can
> also be used for RELOAD, which is probably not a problem because it will
> miss the SAN part (reload URI) and will be rejected.  The problem is that
> any certificate granted for RELOAD can also been used for the Web API, e.g.
> used to modify the configuration file, and letting any user do this is
> probably not a good idea. Adding an EKU in the RELOAD certificates permits
> the enrollment & configuration servers to reject a connection using these
> certificates, without having to use a separate CA.
> 
> A specialized web server could be modified to reject certificates
> containing a reload URI, but because standard web servers already reject an
> EKU that they do not understand, it is a better solution to simply define
> an EKU for RELOAD.
> 
>>>> Regards, André
>>>> 
>>>> Am 24.08.2012 20:09, schrieb Marc Petit-Huguenin: In summary, there
>>>> is a use case for RELOAD that would benefit from having an Extended
>>>> Key Usage (RFC 5280) defined and enforced by RELOAD implementations.
>>>> 
>>>> Here's the use case:  An enrollment server as defined by section
>>>> 11.3 generates certificates that are used by a RELOAD implementation
>>>> as client and server certificates for the purpose of establishing
>>>> (D)TLS links, and to sign RELOAD messages and data.  An enrollment
>>>> server will also probably need an OAM interface, e.g. to provision
>>>> the username/password, and one natural way to design this API is by
>>>> using a RESTful API over HTTPS. Obviously access to this API must be
>>>> restricted and one way of doing this is by using client certificate.
>>>> Buying a certificate that can be used to sign client certificates is
>>>> expensive, so an alternate way is to use a self-signed certificate as
>>>> CA for this purpose (the server certificate still need to be signed
>>>> by a trusted CA), which mean that there is two self-signed CA
>>>> certificates per overlay, one for signing RELOAD certificates, and a
>>>> second one for signing HTTPS certificates.  Having a RELOAD Extended
>>>> Key Usage added in each RELOAD certificate and having all RELOAD
>>>> implementations verifying that the certificates used contain this 
>>>> usage will permit to use the same self-signed CA certificate for
>>>> signing both RELOAD certificates and HTTPS certificates.
>>>> 
>>>> Reducing the number of self-signed certificates from two to one does
>>>> not seem that useful until one is developing an application that
>>>> require to allocate a different overlay to each user.
>>>> 
>>>> There is one use case where enforcing a RELOAD Extended Key Usage can
>>>> be a problem:  There is usages of RELOAD where an Overlay Link Layer
>>>> using Websocket over HTTPS could be useful, in which case an existing
>>>> Websocket implementation would probably reject a connection that is
>>>> using a RELOAD certificate.  That can easily be solved by using code
>>>> that is specific to RELOAD.
>>>> 
>>>> So is there some support to have something this?  If yes, should it
>>>> be added to -base, or developed as an extension?
>>>> 
> -- Marc Petit-Huguenin Email: [email protected] Blog:
> http://blog.marc.petit-huguenin.org Profile:
> http://www.linkedin.com/in/petithug
> 

- -- 
Marc Petit-Huguenin
Email: [email protected]
Blog: http://blog.marc.petit-huguenin.org
Profile: http://www.linkedin.com/in/petithug
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
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=wphE
-----END PGP SIGNATURE-----
_______________________________________________
P2PSIP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/p2psip

Reply via email to