W3C home > Mailing lists > Public > public-ws-resource-access@w3.org > May 2009

RE: [Bug 6403] Enumeration: define policy

From: Geoff Bullen <Geoff.Bullen@microsoft.com>
Date: Tue, 12 May 2009 11:19:54 -0700
To: Doug Davis <dug@us.ibm.com>, "public-ws-resource-access@w3.org" <public-ws-resource-access@w3.org>
Message-ID: <5AAAA6322448AA41840FC4563A30D6E847585C6416@NA-EXMSG-C122.redmond.corp.microsoft.com>
Hi Doug,

On your first point,
What seems to be being said here is that if a new YPath filtering standard comes along, and someone wants to use it with Enumeration, they will have to define the Policy Assertion, and potentially standardize it, if interop is an issue.  While, with the method proposed below, there is an "implied" policy because you can use:

wsen:FilterDialect uri=".../YPath"/>

This seems like a very dangerous assumption, in that there is no real definition for what is meant by using YPath as a filter dialect in Enumeration - it is just "assumed" to work the "way you think it should work".  It seems that is would not be too hard to find examples of where two different implementations of the same filtering mechanism would produce different results under such circumstances.  It would be much safer to actually define (somewhere) what the interaction is between YPath and Enumeration and in the same place define the policy assertion associated with it.  Note that the YPath policy assertion could have extra extension attributes, where as the method below does not allow that (where would they be defined?).

On your second point,
This also ignores the possibility of extension attributes on policy assertions - all policy assertions would have to look and behave the same in order for the user interface to behave in such a generalized manner.  Do you really think that users are going to type in raw filter expressions, possibly in XML, with no syntax checking on the client side?  Is this a real use case?


From: public-ws-resource-access-request@w3.org [mailto:public-ws-resource-access-request@w3.org] On Behalf Of Doug Davis
Sent: Tuesday, May 12, 2009 9:37 AM
To: public-ws-resource-access@w3.org
Subject: RE: [Bug 6403] Enumeration: define policy

Hi Geoff,
  This is actually really close to my proposal - it just changes the NS and localName of the QName - however, there are a couple of aspects of this solution that prevented me from picking this approach in my original proposal.

First, this will probably require the definition of a 3rd spec for each filter dialect.  Let's take the case of XPath.  Let's assume that WS-Enumeration were already completed and behind us (and XPath wasn't known, as filter, at that time).  If someone wanted to use XPath (which has no knowledge of WS-Enum) with WS-Enum, then they would need to define (and standardize) this new QName.  This would require a new spec and its not clear to me who would take on this responsibility. If it didn't come through some org, like w3c, then do we run the risk of each domain creating their own QName for the same dialect?  This could hurt interop, cause confusion and cause us more work.

Second, I'm wondering if we lose the ability for smart clients to be more dynamic because this requires the client to know about all QNames in advance.  When parsing the Policy the client would need to know which QNames were Filter dialects and which were "something else" (which btw it can ignore based on the wsp:Optional attribute). So, if the client wanted to present the user with a drop-down box and a list of filter dialects to choose from, they could only do it for Filters dialects that were hard coded into the system.  Without some way to know that some new QName were a Filter dialect, they would be a bit stuck.  In cases where the client system itself will need to act on these QNames I think its fair to assume that the client must know about each one - however, with Filter dialects the client doesn't actually need to act on them.  Its the service that does all of the work.  So, in a dynamic environment (like the drop-down case I mentioned above) its possible to prompt the user with a list of Filter dialect and an entry field (where they can type the expression) and the subscriber's client code never actually needs to know anything about the Dialect.

This is why ideally I really wanted something like:
        <wsen:FilterDialect uri="...."/>
But of course this would require domain specific policy processing - which I don't think anyone wants.  :-)

I think the mechanism in my proposal, while a bit unique, does provide the same functionality of your proposal but also covers the concerns I mentioned above.
And, it shouldn't cost anything more w.r.t. code since in your proposal you're asking the client to know about the QNames in advance, and there's no reason why they couldn't know about the QNames using the form I proposed.  It just saves the trouble of having to create a new (3rd) spec to link WS-Enum with some expression language.

STSM |  Standards Architect  |  IBM Software Group
(919) 254-6905  |  IBM 444-6905  |  dug@us.ibm.com
The more I'm around some people, the more I like my dog.

Geoff Bullen <Geoff.Bullen@microsoft.com>

05/12/2009 11:51 AM


Doug Davis/Raleigh/IBM@IBMUS


"public-ws-resource-access@w3.org" <public-ws-resource-access@w3.org>, "public-ws-resource-access-request@w3.org" <public-ws-resource-access-request@w3.org>


RE: [Bug 6403] Enumeration: define policy

We propose the following as clarifying amendments to Doug's proposal for Issue 6403:
The Policy assertion for the XPath filter dialect

<wsen:XPathFilterDialect wsp:Optional? />

The policy assertion represents a requirement to include an XPath filter in the WS-Enumeration implementation found at this end point and, specifically, that the attribute: [Body]/wsen:Enumerate/wsen:Filter/@Dialect=" http://www.w3.org/TR/1999/REC-xpath-19991116 " MUST be specified in wsen:Enumerate messages sent to this Web Service.


Per Web Services Policy [WS-Policy]<http://www.w3.org/TR/soap12-mtom-policy/#WS-Policy>, this is compact notation for two policy alternatives, one with and one without the assertion. This indicates that the behavior indicated by the assertion is optional, specifically that a message without an filter dialect is also supported by the endpoint.
This is an extensibility mechanism to allow additional attributes to be added to the element.

The XPathFilterDialect policy assertion element information item MUST NOT include the  wsp:Ignorable attribute in its [attributes] property.
A policy expression containing the XPathFilterDialect policy assertion MUST, if present be attached to either a wsdl:binding/wsdl11:binding or wsdl:endpoint/wsdl11:port.
The normative outline for the Enumeration assertion would be:
<wsen:WSEnumeration [wsp:Optional="true"]? ...>
    <wsen:XPathFilterDialect [wsp:Optional="true"]? />

From: public-ws-resource-access-request@w3.org [mailto:public-ws-resource-access-request@w3.org] On Behalf Of Doug Davis
Sent: Monday, April 27, 2009 10:07 AM
To: Geoff Bullen
Cc: public-ws-resource-access@w3.org; public-ws-resource-access-request@w3.org
Subject: RE: [Bug 6403] Enumeration: define policy

Geoff wrote:
> In general, we are OK with your proposal for Issue 6403.  Here are
> our general comments on the proposal.  I will be happy to provide
> more concrete suggestions in a subsequent email as we are heads-down
> processing multiple WS-RA issues at the moment.
> A)   The <x:FilterDialect> is not a concrete policy assertion but a
> template. This is the first of a kind. How can the WS-RA WG provide
> an XML Schema definition to specify the syntax of the assertion?
> Why not just define a policy assertion for the XPath filter dialect?

Can you give an example of what you're thinking of?  I think its important
to do the policy matching w/o needing domain specific code so I'd be
curious to see what you have in mind.

> We do not understand what is the justification of the statement:
>    "the namespace of this element is application defined, but the
> Local Name MUST be "FilterDialect""
> Why is element matching not using namespaces here?

It _does_ use element matching - so I'm not following.

> >/wsenp:WSEnumeration/wsp:Policy/x:FilterDialect@wsp:Optional
>   This attribute specifies that the assertion is optional per WS-
> Policy. This attribute MUST be present
> B)   The proposal mandates the use of wsp:Optional attribute. But
> the minimum here would be to allow the use of the attribute and
> defer the usage of it to service providers.

We can change this, but my thinking was that the use of the Filter
element and any particular filter dialect (even if just one is
supported) is optional for the client.  But we can change this
if people want and assume people are going to be smart enough
to know that they should really include it - esp. when they support
more than one dialect.

> C)   Would it be useful to check if the proposed assertion follows
> the best practices outlined in the 'Web Services Policy 1.5 -
> Guidelines for Policy Assertion Authors' doc [1].
> At a glance, it appears that the proposal does not follow best
> practices 6, 8, 9, 15, 20, 29 and 31.

Feel free to do so, but the only one that seems interesting is
29 and I think I got that covered, but we can use more WSDL-specific
wording if needed.

> --Geoff
> [1] http://www.w3.org/TR/2007/NOTE-ws-policy-
> guidelines-20071112/#best-practices-list
> -----Original Message-----
> From: public-ws-resource-access-notifications-request@w3.org
> [mailto:public-ws-resource-access-notifications-request@w3.org] On
> Behalf Of bugzilla@wiggum.w3.org
> Sent: Tuesday, April 21, 2009 1:39 PM
> To: public-ws-resource-access-notifications@w3.org
> Subject: [Bug 6403] Enumeration: define policy
> http://www.w3.org/Bugs/Public/show_bug.cgi?id=6403
> --- Comment #4 from Robert Freund <bob@freunds.com>  2009-04-21 20:38:56 ---
> modified proposal as above with the following:
> insert after :/wsenp:WSEnumeration/wsp:Policy/x:FilterDialect
> An endpoint should include a filterdialect policy assertion for each of the
> filter dialects that it supports.
> --
> Configure bugmail: http://www.w3.org/Bugs/Public/userprefs.cgi?tab=email
> ------- You are receiving this mail because: -------
> You are the QA contact for the bug.
Received on Tuesday, 12 May 2009 18:21:41 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:34:49 UTC