#1 implies that r-cards require a WS-Trust STS. This seems to be
a high bar for adoption, particularly when there are several use cases where
the resource UDI is effectively static, and would only ever need to be
retrieved from the STS once. Standing up and maintaining an STS to issue a
one-time claim seems like overkill. m-cards already specify their STS endpoints;
it seems reasonable to treat the r-card endpoint in the same way, no?
From:
higgins-dev-bounces@xxxxxxxxxxx [mailto:higgins-dev-bounces@xxxxxxxxxxx] On
Behalf Of Paul Trevithick
Sent: Sunday, May 10, 2009 9:12 PM
To: higgins-dev
Subject: [higgins-dev] Three proposed changes Higgins R-Cards
For
discussion:
1. Removing the static resource-udi element [1]
After discussion with John Bradley and Markus Sabadello I now see that it would
be better to eliminate the static resource-udi element from the r-card card
XML. Why? (a) It is redundant with the resource-udi claim (b) worse than
redundant, it is static whereas the claim value is dynamic thus it may have an
inconsistent value compared to the dynamic value (c) by removing it an
r-card becomes really just a “profile” of a managed card—the
card .crd file contains no extra XML elements when compared to a regular m-card
(just a rather special claim type URI). The disadvantage is that now any client
(including a selector) must authenticate to the STS and request the
resource-udi in order to learn its value. If it had to do this every time,
performance could suffer. Nevertheless, we’re going to assume that some
clever caching can address this drawback and thus make this change.
2. [Optionally] allowing the resource-udi's value to be an XRD
After discussions with John, Markus and Drummond Reed we now believe that the
value should either be a UDI [as it is today] or [and this is the new option]
an "inline" XRD document. The option to inline the XRD allows what
John calls “private discovery” option in addition to the current [public]
discovery options used during UDI resolution. This addresses some privacy
leakage inherent in public discovery.
Note that there is an interop issue with CardSpace 1.0 & 1.5 in that these
versions (unlike the forthcoming "Geneva" (2.0) release) are badly
behaved WRT XML that they don't understand; namely, that these versions don't
preserve it on export. But the interop situation with this proposed change is
no worse than it was WRT to the previous approach.
3. Renaming "resource-udi" to "resource-udr" at the ICF
Since per #2 above the value isn't necessarily a UDI (it might be a UDI or an
inlined-XRD), we should rename this claim if possible. This should be okay with
the ICF since (a) it was only voted "provisional" [to capture its
“experimental” status] and (b) only one project is yet using this
claim type.
[1] http://wiki.eclipse.org/R-Card#XML_Format
--Paul
