cxf-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dan Diephouse <>
Subject Re: Automatic publication of ws-policy expressions
Date Sat, 22 Dec 2007 07:49:36 GMT
Sergey Beryozkin wrote:
> Hi
> A number of CXF users have come across some limitations of its policy engine which prevent
them from
> meeting otherwise expected results.
> Particularly, what users expect from WS-Policy expressions is to set them up on the server
side and have client runtimes reuse them as appropriate.
> Two issues are on the top of the list.
> 1. Policy engine needs to be explicitly enabled - this one should be simple to fix
> 2. Policies do not automatically get published
> There're two cases here.
> 2.a Java-first development
> 2.b Contract-first, WSDL is already there, policy are defined elsewehere
> For the purpose of the publication policy expressions I'd like to consider 2 cases be
> In both case an issue of privacy may arise, that is, is a given policy expression is
safe to be published ?
> When discussinf WS-SecurityPolicy, I thought we agreed in principle that one way to solve
the issue of privacy is to
> not put the sensitive configuration into the policy expressions but into features and
then the runtime would merge the information appropriately. Thus the WSDL Publisher would
not be concerned about leaking some sensitive data.
> Another approach would be to mark sensitive policy expressions with an attribute like
'private'. There was a concern expressed about solutions like this one. 
> As far as the actual publication is concerned, I thought it would be a matter of policy
components registering themselves as extensors with given WSDL nodes like wsdl:service, wsdl:service/wsdl:ports,
> Thoughts ?

I think I agree that we should out attach to the WSDL. We should have 
some sort of blacklisting mechanism though for policy expressions which 
are private. By default, we should never allow publishing of security 
info (the user shouldn't have to set private=false, it should just never 
show). We should also allow the private=false mechanism.

Have you started work on WS-SecPol? I'm still wishing I had some cycles 
to devote to this...

- Dan

Dan Diephouse
MuleSource |

View raw message