RE: [xri] Difference between global and local subsegments (was RE: [xri]

Steve,

Perhaps there was some misunderstanding. The resolution flows I depict always “honor sticky stars”. What I was pointing out is that it is up to the local policy of the XRI authority answering a resolution request for a global-xref as to how it determines its response.

In other words, in resolving @ootao+west*steve, @ootao is queried for the global xref “+west*steve”. @ootao will *always* answer this request with a single XRDS document containing a single XRD for “+west*steve”.

The only question is how @ootao will figure out its response. One obvious choice is to just register “+west*steve” directly and serve up the XRD for it. Another choice would be for @ootao to *internally* (invisibly to the original resolution request) query itself for +west, then query @ootao+west for *steve, THEN return the response to the original resolver.

And as Marty points out, yet another choice would be for @ootao to query ANOTHER external authority that it recognizes as the authority for “+west”, and then based on that response query @ootao+west for *steve, and then return the response to the original resolver.

Hope this helps,

=Drummond

From: Steven Churchill [mailto:steven.churchill@xdi.org]
Sent: Monday, April 30, 2007 9:21 PM
To: 'Drummond Reed'; 'Schleiff, Marty'; 'Barnhill, William'; xri@lists.oasis-open.org
Subject: RE: [xri] Difference between global and local subsegments (was RE: [xri] Increased complexity in resolver.)

Drummond,

See my single comment in ###.

~ Steve

From: Drummond Reed [mailto:drummond.reed@cordance.net]
Sent: Sunday, April 29, 2007 11:40 PM
To: 'Steven Churchill'; 'Schleiff, Marty'; 'Barnhill, William'; xri@lists.oasis-open.org
Subject: [xri] Difference between global and local subsegments (was RE: [xri] Increased complexity in resolver.)

Steve, Marty, Bill, et al:

It’s Sunday night, and after thinking about Steve’s observation over the last few days, I’ve come to a greater appreciation of the difference between global and local subsegments in an XRI authority segment. (Note that by “global subsegment” I mean the global-literal and global-xref rules on http://wiki.oasis-open.org/xri/XriCd02/XriAbnf2dot1, and by “local subsegment” I mean the local-literal and local-xref rules on that page.)

First, in XRI 2.0, an XRI authority segment consisted of only one global subsegment. It could contain any number of local subsegments inside it (and those could be either local-literals or local-xrefs), but it could not contain another global subsegment.

In the proposed XRI 2.1 syntax, an XRI authority segment can contain more than one global subsegment.

So what Steve noticed is that while the XRI resolution rules are the same for both 2.0 and 2.1, i.e., that a resolver just “walks the tree” of top-level subsegments in an XRI authority segment to resolve it, that tree of subsegments has one important structural difference: once you hit the first global-xref, every other top-level subsegment must be a global-xref.

In other words, think of the pattern this way:

1) XRI authority segments in XRI 2.0 (with the exception of cross-reference root authorities):

GCS-char local local local … /

2) XRI authority segments in XRI 2.1:

GCS-char local local local … global global global … /

I think this makes it easier to see that when you “switch over” from resolving local subsegments to resolving global subsegments, each global subsegment is relative to the previous one just like each local subsegment is relative to the previous one.

Secondly, when you apply that to Steve’s example – of resolving @ootao*west*steve and @ootao+west*steve, it’s true that the first one parses into four top-level subsegments…

ootao

*west

*steve

…and the second one into three…

ootao

+west*steve

However, if the policy of @ootao is that *west and +west are synonyms, then @ootao*west can delegate to *steve and the same delegation can be recognized for @ootao+west*steve. Here’s the flow:

@ootao*west*steve

1) Resolver queries @ for ootao

2) @ responds for ootao

3) Resolver queries @ootao for *west

4) @ootao responds for @ootao*west

5) Resolver queries @ootao*west for *steve

6) @ootao*west responds for @ootao*west*steve

@ootao+west*steve

1) Resolver queries @ for ootao

2) @ responds for ootao

3) Resolver queries @ootao for +west*steve

4) @ootao queries @ootao+west for *steve

5) @ootao+west responds for @ootao+west*steve

6) @ootao responds @ootao+west*steve

### If you follow this to its logical conclusion, it nets out as: it is left to the policy of the authority resolution service as to whether or not it honors sticky stars. Allowing a compliant authority resolution service this type of flexibility will lead to massive interop problems.

Notice that it takes the same number of steps, for the same number of delegations, however @ootao is responsible for answering the +west*steve response, vs. the original client resolver being responsible for resolving it.

This ability to control who will provide the resolution response is, I believe, one of those key guidelines Marty is looking for (and which I am indeed tasked to elucidate) as to whether an XRI author should use local subsegments or global subsegments.

=Drummond (Note: I’m headed off early tomorrow morning for the Higgins f2f meeting in Austin, so I’ll be offline until mid-afternoon tomorrow.)

From: Steven Churchill [mailto:steven.churchill@xdi.org]
Sent: Sunday, April 29, 2007 10:12 PM
To: 'Schleiff, Marty'; 'Barnhill, William'; xri@lists.oasis-open.org
Subject: RE: [xri] Increased complexity in resolver.

> I know that Steve and I lost the "direct concatenization" vs. "compact syntax" vote, but I'd just

> like to point out that under compact syntax "@ootao+west" normalizes to "@ootao*(+west)".

> And if "@ootao*west" and "@ootao+west" are declared as synonyms, then you could logically

> deduce that "@ootao*west*steve" and "@ootao*(+west)*steve" are synonyms.

I agree. I’d just like to point out that the way that the two would be “declared as synonyms” is that “*(+west)” would be added as a local synonym to “*west”.

~ Steve

From: Schleiff, Marty [mailto:marty.schleiff@boeing.com]
Sent: Saturday, April 28, 2007 3:54 PM
To: Barnhill, William; xri@lists.oasis-open.org
Subject: RE: [xri] Increased complexity in resolver.

Hi Bill & Steve (& All),

I think Steve meant that EVEN IF "@ootao*west" and "@ootao+west" are declared as synonyms, then "@ootao*west*steve" and "@ootao+west*steve" are not synonyms (unless they are explicitly declared as synonyms).

I know that Steve and I lost the "direct concatenization" vs. "compact syntax" vote, but I'd just like to point out that under compact syntax "@ootao+west" normalizes to "@ootao*(+west)". And if "@ootao*west" and "@ootao+west" are declared as synonyms, then you could logically deduce that "@ootao*west*steve" and "@ootao*(+west)*steve" are synonyms.

Marty.Schleiff@boeing.com; CISSP
Associate Technical Fellow - Cyber Identity Specialist
Computing Security Infrastructure
(206) 679-5933

From: Barnhill, William [mailto:barnhill_william@bah.com]
Sent: Saturday, April 28, 2007 9:36 AM
To: xri@lists.oasis-open.org
Subject: RE: [xri] Increased complexity in resolver.

Hi all,

Having just gotten back myself I also am trying to catch up.

Steve, can you explain what you mean by "even though
@ootao*west and @ootao+west are both synonyms."?

I thought the resolution sequence of subsegments to be the following after normalization:
@ootao*west => @, *ootao, *west
@ootao+west => @, *ootao, *(+west)
@ootao*(+west) => @, *ootao, *(+west)
@ootao(+west) => @, *ootao, *(+west)

Is this incorrect?

From what you said I got the idea that it's actually
@ootao*west => @, *ootao, *west
@ootao+west => @, *ootao, *west
@ootao*(+west) => @, *ootao, *(+west)
@ootao(+west) => @, *ootao, *(+west)

Which doesn't seem to make sense to me, so I'm betting I misunderstand.

Thanks,
Bill
-----Original Message-----
From: Schleiff, Marty [mailto:marty.schleiff@boeing.com]
Sent: Fri 4/27/2007 11:31 PM
To: steven.churchill@xdi.org; gabe.wachob@amsoft.net; xri@lists.oasis-open.org
Subject: Re: [xri] Increased complexity in resolver.

Hi All,

I'm just trying to catch up on this thread - I've been unavailable all week until now.

I think Steve is right. You can't have a top level local ref following a global ref, because it would end up being part of the global ref.

Even using parens, the closest you could come would be to have a local xref containing a global ref, with a local ref following the closing paren.

I don't envy Drummond for his task of providing guidelines describing when a minter should use local vs. global delimiters. I'll be amazed if he can do it (note that Drummond has amazed me more than once).
--------------------------
Sent from my BlackBerry Wireless Handheld

-----Original Message-----
From: Steven Churchill <steven.churchill@xdi.org>
To: 'Gabe Wachob' <gabe.wachob@amsoft.net>; xri@lists.oasis-open.org <xri@lists.oasis-open.org>
Sent: Fri Apr 27 18:17:48 2007
Subject: RE: [xri] Increased complexity in resolver.

Gabe,

Here's how the proposal would work I think: an authority that has a global
xref as a synonym may have local delegation beneath it (as in node C in my
diagram) but one cannot resolve "through" the global xref. That's why I say
that @ootao*west*steve is not a synonym for @ootao+west*steve even though
@ootao*west and @ootao+west are both synonyms.

Thinking about it more, perhaps I was incorrect in my previous email: maybe
the resolver doesn't need to enforce the restriction after all, because the
parser's syntax tree cannot have a top-level local subsegment following a
top-level global xref.

Drummond, is this last statement true, or is there any way -- using parens
or what have you -- to do this?

~ Steve

-----Original Message-----
From: Gabe Wachob [mailto:gabe.wachob@amsoft.net]
Sent: Friday, April 27, 2007 5:48 PM
To: 'Steven Churchill'; xri@lists.oasis-open.org
Subject: RE: [xri] Increased complexity in resolver.

Steve-

Huh? Maybe because its Friday at almost 6pm, but I'm totally lost.

Why does the next sub-segment have to be a global xref?

-Gabe

> -----Original Message-----
> From: Steven Churchill [mailto:steven.churchill@xdi.org]
> Sent: Friday, April 27, 2007 5:22 PM
> To: xri@lists.oasis-open.org
> Subject: [xri] Increased complexity in resolver.
>
>
> (Please refer to increased-complexity-in-authority-graph.doc. This is the
> same one I sent earlier.)
>
> So the resolver is walking across the top-level subsegments. It encounters
> a
> global xref, so it duly invokes the authority resolution service (of the
> previous subsegment). It gets back an XRD. Now the resolver needs to
> evaluate the type of the next subsegment -- if it's another global xref
> then
> it can continue walking. Else it has to error out.
>
> I would sure hate to see that type of logic in the resolution spec.
>
> ~ Steve
>
>
>
> -----Original Message-----
> From: Steven Churchill [mailto:steven.churchill@xdi.org]
> Sent: Wednesday, April 25, 2007 3:33 PM
> To: 'xri@lists.oasis-open.org'
> Subject: FW: [xri] Flat structure for local delegation within global xrefs
>
>
> I think I've managed to confused everyone (again) by sending the wrong
> document.
>
> The attachment increased-complexity-in-authority-graph.doc summarizes my
> discussion with Drummond yesterday. For one thing, I fear that we've
> redefined the meaning of the term "XRI synonym". In the diagram you will
> note that when adding the localID "+west", @ootao*west is (still) a
> synonym
> for @ootao+west (they resolve to the same XRD), but @ootao*west*steve
> would
> no longer be a synonym for @ootao+west*steve.
>
> The other attachment (increased-complexity-in-authority-graph-
> flatness.doc)
> talks about the flat structure for local delegation within global xrefs.
>
> ~ Steve
>
>
>
> ________________________________________
> From: Steven Churchill [mailto:steven.churchill@xdi.org]
> Sent: Wednesday, April 25, 2007 3:22 PM
> To: xri@lists.oasis-open.org
> Subject: [xri] Flat structure for local delegation within global xrefs
>
> (Blech.)
>
> Please see attached.
>
> ~ Steve
>

xri message