> would mean that @free*c, @free*d, @free*a*b and @free*x*y*z would
all have the same CanonicalID, which of course would break
> CanonicalID Verification.
XRI Refs allow these identifiers to resolve to the same CanonicalID.
From: Markus Sabadello
[mailto:markus.sabadello@gmail.com]
Sent: Wednesday, May 16, 2007 5:37
PM
To: Steven Churchill
Cc:
openxri-users@lists.sourceforge.net; andy.dale@ootao.com;
xri@lists.oasis-open.org
Subject: Re: [Openxri-users] [xri]
Proposed draft text for WD11 introducingthe authority graph model. [Was
Thoughts about XRI aliases.]
For those four aliases, there simply is no "absolute identity". It's
just an internal behavior of OpenXRI. Technically speaking, the four i-names
share the same database rows, that's all. There can be 1000 other XRIs out
there that also resolve to www.inames.net
or even have the same SEP set as my four i-names, and I would still not
consider them "aliases" under my OpenXRI-specific terms.
Since they (currently) don't even have CanonicalIDs, I would simply say there
cannot exist any global abstract model that could define them as synonyms in
the way I mean it when I say "aliases". But who cares?
Once OpenXRI has CanonicalIDs this will be different. However, I am not yet
sure how. One possibility would be to put absolute CanonicalIDs into the
store.. This would mean that @free*c, @free*d, @free*a*b and @free*x*y*z would
all have the same CanonicalID, which of course would break CanonicalID
Verification. Another possibility would be to store what you call local
canonical identifiers. Then the four "aliases" would have verifiable,
but different CanonicalIDs (except @free*c and @free*d; they would still have
the same).
I guess both is not perfect....
-Markus
On 5/17/07, Steven
Churchill <steven.churchill@xdi.org
> wrote:
Markus,
Thanks for your response.
Here's something you might consider. You say that the
following are aliases:
- @free*c
- @free*d
- @free*a*b
- @free*x*y*z
as I said in the first paragraph of my proposed text: "Identifier
synonymity cannot exist outside of an abstract model that formalizes the
absolute identity of the object for which two identifiers purport to be
synonyms." For example, one cannot say that "Steven
Churchill" and "Steve Churchill" are synonyms for "me"
(the guy typing this) without first having a model in which the absolute
identity of "me" is defined.
So if the above XRIs are all aliases, then we should examine the absolute
identity to which they are all aliases. It seems that, in your case, your
model's notion of absolute identity is "the XRI authority that returns an
SEP containing URI http://www.inames.net under
default service selection." (Or perhaps it is: "the XRI authority
that returns this particular set of SEPs containing these values".
XRI resolution certainly does not preclude a synonym/identity
model that defines the identity of an authority in any of these ways. But the
more rigorous the notion of identity, the more practical application you will
get from your model.
~ Steve
This is very good. Something like that would have helped me a lot to understand
XRI resolution in the beginning. I especially like this paragraph:
"XRI Resolution does not preclude alternative models of XRI synonymity
and/or object identity, however this specification provides no mechanism for
synonym verification under such alternative models."
This would legitimize OpenXRI's current approach to aliases, which I really
would like to keep (and Gabe too I believe). I think I keep calling them
aliases, because I instinctively felt that they were not quite compatible with
the synonym mechanisms defined in the Resolution Spec.
I now actually set up the example I talked about in my other mail a few days
ago. I configured the following XRIs as aliases "inside" the OpenXRI
server:
- @free*c
- @free*d
- @free*a*b
- @free*x*y*z
They share all SEPs and should resolve to http://www.inames.net. For *c and *d, appropriate LocalIDs
are automatically included in the XRD.
By the way, of course I agree with you that the OpenXRI server would be pretty
useless without CanonicalID support.. As seen here, this is on top of the list
:)
http://dev.inames.net/wiki/OpenXRI:RoadMap#Server_-_Authority
Your definitions 1.6.1, 1.6.2 and 1.6.3 will be helpful to properly implement
this.
-Markus
On
5/15/07, Steven Churchill < steven.churchill@xdi.org>
wrote:
All,
Enough of my complaining with regard to the XRI Resolution
spec neglecting to define its abstract model for the resolution authority
graph.
I've attached a proposal for text to be added to WD11.
I look forward to your comments.
~ Steve
|