RE: [security-services] removal of AttributeValueType still an is sue

Title: RE: [security-services] removal of AttributeValueType still an issue

Apologies to all, because I haven't followed this debate as closely as I should have.  I, personally, see advantage in a scheme that would have attributes in assertion instances actually tagged with the attribute type, rather than tagged simply "attributeValue" and qualified by an xsi-type attribute.  Presumably, a substitution group would achieve this.  Can anyone see disadvantages to this approach?  All the best.  Tim.

-----------------------------------------

Tim Moses

Tel: 613.270.3183

-----Original Message-----

From: RL 'Bob' Morgan [mailto:]

Sent: Friday, January 18, 2002 6:23 PM

To: OASIS Security Services TC

Subject: [security-services] removal of AttributeValueType still an

issue

So, despite the fact that the change below (removing AttributeValueType

and defining AttributeValue as anyType) has been agreed upon multiple

times, core-25 still specifies an AttributeValueType.

It may be that Hal intended issue [DS-4-08: anyAtttribute], which is

marked as open, to describe this issue, but that description doesn't

include the precise fix that Chris proposed in his message, to which all

interested parties appear to have agreed.

I am becoming frustrated by how difficult it has been to get this resolved

and into the core spec (or, discussed if anyone wants to argue against

it).  We're trying to decide on the use of AttributeValues right now in

Shib and we need this schema to be resolved.

 - RL "Bob"

---

On Mon, 7 Jan 2002, RL 'Bob' Morgan wrote:

>

> On Wed, 19 Dec 2001, Eve L. Maler wrote:

>

> > At 12:10 AM 12/19/01 -0400, Chris McLaren wrote:

> > >

> > >My inclination would be to remove the AttributeValue type and set the

> > >type of AttributeValue directly to the anyType. This would removes

> > >nothing functionally from the AttributeValue and allows us to do the

> > >sort of direct xsi:type-ing that I mention in my earlier posts.

> >

> > I agree with this.  In the earlier thread explaining the "trick", I had

> > missed the fact that we don't, yet, define AttributeValue as anyType.

>

> There seemed to be agreement among those who cared about it that this

> change should be made, but as far as I can tell core-22 still defines

> AttributeValueType as a sequence of any.  So, let this be noted as a

> proposed and apparently agreed-on but not yet implemented change to the

> doc, and hence an open issue.

>

>  - RL "Bob"

----------------------------------------------------------------

To subscribe or unsubscribe from this elist use the subscription

manager: <http://lists.oasis-open.org/ob/adm.pl>