RE: [ebxml-msg] Some suggestions regarding default securitysettings in ebMS 3.0

[Sacha Schlegel <sacha_oasis@schlegel.li>]

Hi Jacques

Am Dienstag, den 13.02.2007, 16:46 -0800 schrieb Durand, Jacques R.:
> Sacha:
> 
> If we are absolutely confident that no use case that requires a
> different security order is worth supporting, then the expected order
> can be required as in ebMS2.

I mixed them up in my email as Dale pointed out. So for the records:

As noted in Note in section 4.1.4.5 in ebMS 2.0 sign first then encrypt.

> 
> Otherwise, it seems to me that could be handled in the "gateway"
> conformance profile - i.e. this profile would require support for a
> particular order of security operations.
> 
> Although conformance profiles are not defined in the main spec, it is
> assumed you and partners need to agree on one in order to get
> interoperability. The "gateway" conf profile is supposed to be the
> baseline for eB/eG exchanges.

Will these conformance profiles have standardised forms or structures or
unique ways to reference them? Are they part of ebXML MS or ebXML IIC
TC? 

Would you think that theses conformance profiles will then be referenced
in the ebXML CPA?

In the ebXML CPA team of course we want to have everything ready once we
come to a final ebXML CPA. The idea being that we only need to agree on
the values we put in the ebXML CPA (which can include to set the
conformance profile) and than there is no need to later have yet another
agreement (discussion, email exchange or any other form of
collaboration) to achieve interoperability.

> 
> If we consider this is something variable enough to be a P-Mode
> parameter (i.e. subject to agreement at deployment time, not just at
> development time) then the gateway conformance profile could require
> this parameter to have the value specifying the desired order. So every
> P-Mode derived for the gateway profile would specify this order, even if
> the implementation can support other orders.
> 
> To summarize: such a requirement could appear at 3 levels (from the most
> inflexible to the most flexible):
> (a) core spec (like in V2)
> (b) conformance profile
> (c) P-Mode parameter and value.
> 
> My hunch is that (b) or (b)+(c) could be appropriate.
> 

The way I understood the ebXML CPA phone conference call where these
questions were raised was if I want to use ebMS 3.0 and ebXML CPA 3.0
must an application supporting both implement yet another WS-*
specification (I think WS Policy was referenced) or can we have all ebMS
3.0 configurations done in plain ebXML CPA.

So, Dale Monica please help out if I miss the point here. I think as
long as these configurations values can be configured without the
absolute need to have another WS-* spec implemented (Monicas concern was
as, far as I understand, that that requirement could have further
consequences, eg if you have to implement WS-A for the configuration
only then you naturally would also need implement WS-B, WS-C etc, Monica
maybe you clarify again) then for the ebXML CPA people that would be OK
and it is the ebXML MS people task to find the right place to place
those settings.

Sorry I do not know about the P-Modes enough but your hunch is that they
would not be in the spec.

Hope this makes sense and sorry that I did not keep up with ebMS 3.0
specs.

Sacha

> Jacques
> 
> -----Original Message-----
> From: Sacha Schlegel [mailto:sacha_oasis@schlegel.li] 
> Sent: Friday, February 09, 2007 2:26 PM
> To: ebXML Messaging TC
> Subject: [ebxml-msg] Some suggestions regarding default security
> settings in ebMS 3.0
> 
> Dear ebMS group
> 
> In todays ebXML CPPA conference call Dale talked about the ongoings in
> ebMS 3.0 and what ebCPPA will have to accommodate to align with ebMS
> 3.0.
> 
> One question was formed around "given the diversity of deployment
> environments, what is a easy way to configure an ebMS 3.0 with a minimal
> or optimal set of existing or new technologies?"
> 
> In this discussion, signature and encryption were identified as two key
> functions, and the order in which they occur. It was noted that ebMS 3.0
> no longer specifies the default configuration as was defined in ebMS
> 2.0.
> 
> ebMS 2.0 has two defaults:
> 
> a) encrypt first, then sign. As a Note in section 4.1.4.5
> 
> b) the Reference for the actual ebMS 2.0 SOAP message XML digital
> signature was set in Section 4.1.3. The Reference 
> 
> <Signature>
>  <SignedInfo>
>   ...
>  <Reference URI="">
>   <Transforms>
>     <Transform
> Algorithm="http://www.w3.org/2000/09/xmldsig#enveloped-signature"/>
>     <Transform Algorithm="http://www.w3.org/TR/1999/REC-xpath-19991116";>
> 
> <XPath>not(ancestor-or-self::node()[@soap:actor="urn:oasis:names:tc:ebxm
> l-msg:actor:next=MSH"] |
> ancestor-or-self::node()[@soap:actor="http://schemas.xmlsoap.org/soap/ac
> tor/next"])</XPath>
>      </Transform>
>    </Transforms>
>   </Reference>
>  </SignedInfo>
> </Signature>
> 
> Monica worded the context of all this nicely:
> 
> Without such defaults in ebMS 3.0, several concerns come into play.
> Where does the onus fall to configure, and what implications does this
> have to existing or future implementations? One scenario identified was
> that WS-Policy and the domain specific WS-Security Policy could apply in
> the CPA 3.0. Yet, did ebMS 3.0 intend to implicitly extend dependencies
> to even more specifications to provide simple (and flexible to complex)
> mechanisms to support configuration of, for example, P-Mode parameters? 
> 
> The simpler approach would be for ebMS 3.0 to retain some default
> specification as existed in ebMS 2.0. A simple default allows
> implementations to transition functionality in an incremental way,
> particularly as the community gains more experience with these
> technologies (including those cited above). It makes sense to consider
> defaults to facilitate that transition in specifications and in
> technology adoption.
> 
> Pete, who was also on the phone conference call, added that he thinks
> WS-Policy lacks the operation order (sign or encrypt first) as well as
> properly dealing with attachment references. Pete also mentioned that
> such default values matter on aspects such as non-repudiation,
> transitivity, tamper evident messages etc (Pete may have to explain
> better himself) so it is no easy choice how such defaults are defined.
> 
> Kind regards
> 
> Sacha Schlegel
>