[Idschemas] OpenID and Higgins attribute harmonization

[Recordon, David]

I'm +1 to this plan! Dick (and his team) however knows best as to the
changes needed to AX. As this data would conceivably be cached anyway,
adding an additional fetch for cases B and C (well also following
redirects I assume) does not bother me.

--David


________________________________

From: idschemas-bounces at idcommons.net
[mailto:idschemas-bounces at idcommons.net] On Behalf Of Paul Trevithick
Sent: Monday, January 08, 2007 7:58 AM
To: idschemas at idcommons.net
Subject: [Idschemas] OpenID and Higgins attribute harmonization



[We had some problems with our mailman server over the weekend so you
may have already received this email]

 

Over the holidays I chatted with Dick about how to harmonize his OpenID
Attribute Exchange (AX) proposals for OpenID with the Higgins approach. 

 

There are other conversations going on with Drummond on how to leverage
XRI dictionaries. And there are related conversations with Microsoft and
IBM and others about how to extend the CardSpace schema. But let me here
just bite off one chunk....

 

WRT to just OpenID AX and Higgins, here's where we are:

 

- We agree that URIs should be used to describe attribute types (in
higgins.owl these are higgins:attribute sub-properties) as does
MSFT/CardSpace

 

- We agree that we need to associate metadata with these URIs. Such
metadata would allow a system to learn about the attribute and might
allow it to "map" it to some equivalent or nearly equivalent attribute
type that it already knows how to process.

 

We differ on three key points:

 

1) In OpenID AX these URIs are resolvable to an XML document that
describes the attribute whereas in Higgins we do the opposite. In
Higgins we follow the RDF/OWL precedent of having a separate document
that contains metadata "about" these URIs. 

 

Let me repeat this key point in different words. In AX the attribute
type knows how to find the metadata that describes itself. In Higgins
the executing system somehow knows where to find a relevant *.owl file
that describes a set of attributes and that hopefully contains a section
that describes the attribute type URIs that the IdA, RP, or IdP system
encounters.

 

2) AX and Higgins differ as to the semantics of the metadata about each
attribute

 

3) AX and Higgins differ on the document format used to express the
semantics. AX uses XML-Schema/XML, Higgins uses OWL/RDF/XML.

 

Okay, so let's take these in order...

 

WRT 1) above, since I'm personally compelled by the advantages of
self-describing attribute type URI/URLs, I wanted to find a way to
bridge the differing approaches. What occurred to me was that if we
could add an optional extra level of indirection/lookup to AX, we could
make the two approaches work together. Here's what I was thinking:

 

- We allow the attribute type URI to resolve to a document that EITHER
(a) contained directly the metadata about it or (b) contained the URI of
a metadata document or (d) contained the URI of a metadata document and
a section within it. 

 

At present AX only has option (a) above. It would need to add support
for

(b) and (c). Option (b) would allow indirection into a well known
repository of multiple metadata documents. This might be nice to have.
Not sure. Option

(c) is a must have. Option (c) would allow an attribute type URI to
refer to one specific section of a document. The section would describe
the metadata about this particular attribute type, but the document
could also contain sections about other attributes as well (as Higgins
schema descriptions tend to do). 

 

[No doubt the astute XRI enthusiast will jump in here and propose that
the attribute URI could resolve to an XRDS document that would provide
the indirection and that this indirection should be an XRI cuz pointing
"inside"

another document is what XRI's can do in their sleep.] 

 

WRT 2) above we have some work to do to bridge things. (I'm hoping that
Joaquin Miller is still willing to help me with this). What we need to
do is create a three column table: column one is entitled "semantic",
column two is "OpenID AX", column three is "Higgins". We need to show
what semantics are supported in common, and what semantics are unique to
each. [In some cases the "semantic" expresses the allowed syntax of the
attribute's data value, but let's consider them "semantics" all the
same.]

 

For the semantics that AX supports that Higgins does not, Higgins will
almost certainly immediately add them to the Higgins metadata model.
[This will involve Higgins moving from OWL-DL to OWL-FULL, but we've
been teetering on that brink for a while. For example, Novell has some
semantics from LDAP that Higgins will only cleanly be able to model if
we relax our restriction on using OWL-DL.]

 

If Higgins did this (and at a glance I already know that some
non-trivial work is involved for the Higgins group) then there is a
quick way to resolve

3) above....OpenID adopts the Higgins metadata document format used in
options (a), (b) or (c) above.

 

So that's a possible road map for harmonizing our 1), 2) and 3)
differences.

 

-Paul

 

PS: The burst of interest in this schemas topic (i.e. the "profile
exchange"

thread) in the larger Identity Gang list this weekend shows how timely
this all is.