W3C home > Mailing lists > Public > public-webapps@w3.org > January to March 2009

Re: ACTION-306: Trust anchors

From: Frederick Hirsch <frederick.hirsch@nokia.com>
Date: Wed, 25 Feb 2009 17:55:13 -0500
Cc: Frederick Hirsch <frederick.hirsch@nokia.com>, "public-webapps@w3.org WG" <public-webapps@w3.org>
Message-Id: <04239CA6-B30E-4204-8FEF-2D3A7B94F9C0@nokia.com>
To: ext Thomas Roessler <tlr@w3.org>
Thanks for the proposal Thomas.

This proposal requiring Basic Path Validation seems to conflict with  
X509Data being optional, the current language that I think we  
discussed during the meeting:

Generation:
5c) The ds:KeyInfo element MAY be included and MAY include  
certificate, CRL and/or OCSP information. If so, it MUST be compliant  
with the[XMLDSIG11] specification. If certificates are used they MUST  
conform to the mandatory certificate format.

Validation:
If a ds:KeyInfo element is present then it MUST conform to the  
[XMLDSIG11]specification. If present then any certificate chain SHOULD  
be validated and any CRL or OCSP information may be used as  
appropriate [RFC5280]..

I suggest we could also adopt your text by changing the final sentence  
above  to

If present then user agents MUST perform Basic Path
Validation [RFC 5280] on the signing key and SHOULD perform revocation  
checking as appropriate.  The set of acceptable
trust anchors, and policy decisions based on the signer's identity
are established through a security-critical out-of-band mechanism.

Question:
Should re require use of X509Data to convey certificates?

I was suggesting not, since this could be conveyed out of band and it  
might not always be appropriate to include in every signature.

Thoughts on this one?

regards, Frederick

Frederick Hirsch
Nokia



On Feb 25, 2009, at 9:23 AM, ext Thomas Roessler wrote:

> I propose that we add te following text in the beginning of 6.2:
>
>> The validation procedure given in this section describes extensions
>> to XML Signature Core Validation.  In addition to the steps defined
>> in these two specifications, user agents MUST perform Basic Path
>> Validation [RFC 5280] on the signing key.  The set of acceptable
>> trust anchors, and policy decisions based on the signer's identity
>> are established through a security-cirtical out-of-band mechanism.
>
> (If somebody can think of something nicer to say, that's fine as
> well.  Note that the Basic Path Validation requirement isn't really
> new -- it's implicit to our use of X.509, if done properly.
> Nevertheless, worth calling out properly.)
>
> --
> Thomas Roessler, W3C  <tlr@w3.org>
>
>
>
>
>
>
>
>
Received on Wednesday, 25 February 2009 22:56:25 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 18:49:30 GMT