Drummond says “IMHO the authority of an authority
server is sacrosant. It can apply any policy it wants to produce the next XRD.”
Hmmm except for where we have already put policies in place.
For example
= personal
@ organization
+ dictionary
$ system
* reassignable
! persistent
From: Drummond Reed
[mailto:]
Sent: Wednesday, November 26, 2008 7:11 PM
To: Chasen, Les;
Cc: ;
Subject: RE: [xdi] on global cross references and the link contract use
case
What the @cordance authority service does when given the
subsegments…
=drummond
*(=drummond)
…is up to the @cordance authority server. I agree that
@cordance authority server could decide to consult another authority server in
its response, but it still controls the response it its own context no matter
where it gets the XRD or what it puts in it. For example, it could:
1) Have a policy that both of the subsegments above are synonyms in
its context and that they return the same XRD.
2) Have a policy, as Les suggests, that *(=drummond) will return
the XRD (or selected SEPs from XRD) from the = registry. (See note below.)
3) Have any other policy that applies to global subsegments and/or
local cross-references.
IMHO the authority of an authority server is sacrosant. It can
apply any policy it wants to produce the next XRD. What matters is that:
A) Parsing of the next subsegment to be resolved is always 100%
unambiguous for a resolver.
B) The resolver always hands the authority server EXACTLY the next
subsegment to be resolved (or all remaining subsegments to be resolved in
lookahead resolution).
C) The authority server return a valid XRD (or an XRDS in the case
of lookahead resolution).
=Drummond
From: Chasen, Les
[mailto:]
Sent: Wednesday, November 26, 2008 10:03 AM
To:
Cc: ;
Subject: Re: [xdi] on global cross references and the link contract use
case
I
need to stick with simple examples and in this case there is only one
authority, @cordance, in question. under the current proposal @cordance is
asked for both =drummond and *(=drummond) and it returns its own xrd if one
exists within @cordance. We have never discussed whether @cordance may go to
the = authorty server for drummond when it is contained within *(=drummond). I
think that maybe interesting behavior.
--------------------------
http://xri.net/=les.chasen
From: John Bradley
To: Chasen, Les
Cc: ;
Sent: Wed Nov 26 12:44:00 2008
Subject: Re: [xdi] on global cross references and the link contract use
case
Les,
I started off with that position. I now think
they are separate queries to the next authority server.
I think that the parenthesized statement may
contain multiple subsegments is the important thing.
Once we have a way to encapsulate multiple subsegments to be
handed to an authority server it is hard to stop someone from
only putting in a single subsegment. in the cross ref.
I think the simple principle is that things
in parenthesis are opaque to the resolver and handed to the next
authority server and things not in parenthesis are resolved left to
right one subsegment at a time.
I also don't think that in ether case @cordance is being
asked anything about the global =drummond if one exists.
In the first case the authority server for @cordance is
being asked for a XRD for the subsegment =drummond the = is treated as part of
the subsegment itself
In the second case the authority server for @cordance
is being asked for a XRD for the subsegment *(=drummond)
I think the latter could be taken that the authority server
may look someplace else to get the XRD for the cross reference.
I think in the former case =drummond is
a subsegment in its authority server.
The question is if it is just a regular subsegment is using
= as a separator going to confuse people.
=drummond is just a regular subsegment with
some inference by Cordance that its =drummond has something to do
with the global =drummond though XRI makes no such claim other things using XRI
like XDI may.
John B.
On 26-Nov-08, at 9:23 AM, Chasen, Les wrote:
Let's
not complicate this with +phone. My question is does @cordance return a
different xrd for =drummond and *(=drummond)? This proposal says yes. I
disagree with this behavior. I think @cordance is being asked for its
representation of =drummond in both cases.
--------------------------
http://xri.net/=les.chasen
From: John Bradley
To: Chasen, Les
Cc:
;
Sent: Wed Nov 26 12:11:59 2008
Subject: Re: [xdi] on global cross references and the link contract use
case
From a XRI resolution perspective I see
a difference in what @cordance's authority server is asked for.
@cordance=drummond
, @cordance is asked for the XRD for =drummond
@cordance=drummond+phone
, @cordance is asked for the XRD for =drummond
, @cordance=drummond is asked for the XRD for +phone
@cordance*(=drummond)
, @cordance is asked for the XRD for *(=drummond)
@cordance*(=drummond+phone) , @cordance is
asked for the XRD for *(=drummond+phone)
Parenthesis in the first segment tell the resolver to
treat the contents as an opaque string and pass it to the next authority
server.
The resolver Parenthesis in the path are
not significant to the resolver they would
be matched during service selection as they are now.
One thing we did start talking about at the F2F is what the
syntax to indicate special processing on a cross reference.
An example:
$XRD*(https://boing.com)*marty
Is $XRD a node that has as its authority service one that
performs XRD resolution on the next subsegment.
So is a $ word in XRI resolution a node that
points to a specialized authority service?
Certainly a resolver has the option
of shortcutting resolution through querying the $XRD
authority server if it understands the $XRD word.
Under Drummond's proposal we do have to have a theory about
what + and $ in the first subsegment resolve to.
John B.
On 26-Nov-08, at 6:51 AM, Chasen, Les wrote:
Hi
Giovanni -
A cross reference is putting the thing cross referenced inside the context of
the xri that contained the cross reference. Terminology in this proposal has
been changing but the thing that hasn't is that in both cases the outer
authority is returning its xrd for the reference. That is, in the example
below, cordance is returning its xrd for =drummond.
@cordance=drummond
@cordance*(=drummond)
So we only support the 'inside' (or better stated container) concept. Now it
may be interesting to consider a resolution process which rather than asking
@cordance for =drummond we ask = for drummond. Today that does not exist.
Have a happy thanksgiving.
Thanks
Les
--------------------------
http://xri.net/=les.chasen
From: Giovanni Bartolomeo
To: Chasen, Les
Cc: XDI
Sent: Wed Nov 26 09:35:43 2008
Subject: RE: [xdi] on global cross references and the link contract use
case
Hello Les,
First, thank you for having read the contribution! Unfortunately I'm probably
not the best person which can answer your question, as I'm not an expert of
XRI2.0. I limited to develop my contribution in the scope of the current
proposal (
http://wiki.oasis-open.org/xri/XriThree/GcsDelimiter), which, I personally
support for one specific reason: http://lists.oasis-open.org/archives/xri/200811/msg00043.html.
I think Drummond's reply http://lists.oasis-open.org/archives/xdi/200811/msg00031.html
is a good technical explanation of this, expecially the last statement: in
an ordered set, you are not referencing across contexts. The context is the
ordered set itself. For example, in $sig$d, the context for $d is $sig.
Which I understand as follow: put an object inside a context is not exactly the
same as referencing it from that context).
The core of this issue - that's my guess and not a technical explanation
- is probably that XRI2.0 did not consider the need to "put an
instance of an object inside a context", but just to "cross
reference" it. That is instead a fundamental capability for XDI (and for
any human or machine understandable language).
I'll be happy to answer any further questions you might have on the contents of
the contribution itself.
Kind Regards,
Giovanni
At 13.36 25/11/2008, Chasen, Les wrote:
Hi Giovanni. I had a chance to read this. While
I like what you guys
are doing with link contracts I fail to understand why all of these
examples do not work with XRI 2.0 style cross references.
> -----Original Message-----
> From: Giovanni Bartolomeo [
mailto:]
> Sent: Wednesday, November 19, 2008 12:02 PM
> To: XDI
> Subject: [xdi] on global cross references and the link contract use
> case
>
> Hello,
>
> some weeks ago I mentioned that I was playing a bit with the
> semantics of link contracts. Better to say, with the global cross
> references applied to the link contract use case. After the issue of
> XDI RDF model v.11, I put my thoughts in a document which you can find
> below.
> Maybe just semantic nuances, but your comments will be welcome.
>
> Thanks,
> Giovanni
No virus found in this incoming message.
Checked by AVG - http://www.avg.com
Version: 8.0.175 / Virus Database: 270.9.10/1811 - Release Date: 25/11/2008
8.29
