RE: [xdi] about addresses, synonyms and equivalence (was: Re:Agenda: XDI TC Telecon Thursday 1-2:30PM PT 2011-03-24)

["Barnhill, William [USA]" <barnhill_william@bah.com>]

Thanks Giovanni, some responses inline, prefaced with a [wab].


> good questions which could open a long discussion. Anyway please find some first quick thoughts below...

>> Another thought to consider regarding synonyms...
>> Given =abc/$/=!1234, is =abc always resolution equivalent to =!1234,
>> or is this resolution equivalent scoped to the context of the
>> requester?

>To me they are synonyms, see below why I believe this.

>>
>> Example:
>> Let's say I have two personas, =eng and  =bill. The current method
>> reveals the i-number via =bill/$/!1234 when the =bill graph is
>> queried. If the requester has the permissions to see =bill and =eng
>> from the same data authority, but not the link between them, then
>> there's a problem when =eng/$/!1234 is seen, because !1234/$$/=eng is
>> implied and from that =bill/$/=eng, if resolution equivalence is a
>> symmetric relation.

> This is the case if we state that !1234 and =bill are synonyms and
> that =eng and !1234 are synonyms. In my view, =bill, =eng, !1234 (is
> that correct or =!1234 is the correct form?) are all XRI addresses
> which are synonyms, i.e. that resolve to the *same* *node* in the global graph.

> [I deliberately leave open the matter of what is a node and which
> identifier/address it is assigned - to me it depends on the binding
> context (likely http) just as a XRI depends on the binding context to
> turn into a concrete identifier, but this is another topic. When the \
> binding is chosen you can then resolve the XRIs. If they *resolve* to
> the same http address, e.g., they are "resolution equivalent". As before
> stated, synonyms must always resolve to the same node and thus to the
> same http addresses and therefore I call them "resolution equivalent"].

[wab] If you have a node that is available on more than one channel, for example XMPP and HTTP, and two requests resolve the data through the different channels are those channels resolution equivalent? To me they are iff  the data transported over the channel is the same data in the graph, i.e. an update to that data through either channel would result in the same graph transformation.

>> To get around that personas can't be synonyms they have to be
>> sub-contexts (i.e., narrowing of the meaning of the
>> specific individual).  That means =eng/$/=!5678 and =bill/$/=!1234.
>> So then is =eng/$/=bill if you can see the whole graph? I wouldn't
>> think so.
>>

> I agree that synonyms are not at all a way to model different personas.

> About how to model personas in XDI there is a thread started here
> http://lists.oasis-open.org/archives/xdi/201011/msg00024.html and not
> yet closed. Please see also my replies at
> http://lists.oasis-open.org/archives/xdi/201011/msg00030.html and
> http://lists.oasis-open.org/archives/xdi/201011/msg00031.html.
> These clearly show that personas are not treated as synonyms.

[wab] Yet several of the discussions I've seen treat personas as synonyms I believe.  I agree that different personas should not be synonyms of each other and that's part of what I was trying to say, that I think the current graph model may lead implementers down the wrong path of treating personas as synonyms.  I remember those conversations but need to re-read them to refresh.

>> The point I'm trying to make is logical equivalence vs resolution
>> equivalence: should resolution equivalence really be a symmetric
>> relation and defined similar to owl:sameAs, or an asymmetric relation
>> and defined similar to skos:narrows?  I think the skos:narrows
>> approach actually could give us nearly as much semantic
> expressibility, perhaps more if we repurposed a second GCS to indicate
>> narrowing/broadening(when inverted) vs equivalence ?

> Both mechanisms are useful, but in different situations. I've tried to
> explain before how synonyms (owl:sameAs) and "resolution equivalence"
> are bound together. Narrowing and broadening are something which I wouldn't
> call "equivalence" (lacking of the transitive property which is instead in
> any equivalence relationship), but, yes, they could be useful concepts to
> include. As regarding symbols, $= could be a choice for synonymity, provided
> we need one (we could simply leave a xref to owl:sameAs instead...)

[wab] I remember our conversations and agree that what I am talking about would not be equivalence.  I've always maintained the most powerful relations we have are subset-of, member-of, and their inverses.  The existence of $words for those, and the accepted $words to use, has changed over time. Inverse of member-of is currently $has I believe, so member-of is $$has (I still think we should have a $in for that). I've always gone with the approach that an individual is a set of max cardinality 1.  That leads to member-of and subset-of being the same relation. For example 23-mar-2011 is a subset-of each in [23-mar,mar, mar-2011, 2011] and  each of those is a subset-of $d.

>> To me '=' would fall into that well, despite my dislike of the
>> repurposing technique it'll work - if we explain it often and early, I
>> think.
>>
>> =bill/=/=!1234 -> =bill is synonym of of =!1234
>> =eng/=/=!5678 -> =eng is synonym of =!5678 Hidden, unless explicitly
>> revealed via link contract...
>> =!1234/$/=!90 -> =!1234 is a facet of =!90 (the actual person)
>> =!5678/$/=!90 -> =!5678 is a facet of =!90
>>
>> Also makes more sense to me, as my persona as =bill implies more than
>> my persona as an engineer but =bill might be my personal not
>> professional life, which means =eng implies more than =bill as well,
>> so there has to be an over-arching identity gem that encompasses, but
>> is not equal to, all these facets.
>>
>> Kind regards,
>> Bill
>>
 ________________________________________
>>> From: Giovanni Bartolomeo [giovanni.bartolomeo@uniroma2.it]
>>> Sent: Tuesday, March 29, 2011 7:59 AM
>>> To: xdi@lists.oasis-open.org
>>> Subject: [xdi] about addresses, synonyms and equivalence (was: Re:
>>> Agenda: XDI TC       Telecon Thursday 1-2:30PM PT 2011-03-24)
>>>
>>> Dear Drummond,
>>>
>>> let me answer your questions, with some clarifications. Let's start
>>> first from pointing out terminology: "logical equivalent" is probably
>>> not exactly what you describe, which we agreed to call instead
>>> "resolution equivalence", i.e. two addresses resolve to the SAME node
>>> (logical equivalence is a broader concept we discussed in the past,
>>> but for this discussion it is not essential, so I'll skip this part :)
>>>
>>> Coming back to the scenario, =abc and =!1234 are synonyms. As such,
>>> they are two addresses pointing at the SAME node in the graph.
>>> Analogies: two IP addresses assigned to the same host in a network or
>>> two variables pointing at the same java object in java programming. As
>>> such, they both should be resolved to the SAME node, even if they are
>>> syntactically not the same.
>>>
>>> So when you ask
>>>
>>>> So the problem is: which is it? Which of these two addresses should
>>>> the XDI software expect to find? Both of them? Does that means Is
>>>> every subcontext and/or property of every node in the graph that has
>>>> a synonym logically equivalent to having that same subcontext and
>>>> /orproperty on every other synonymous node in the graph?
>>>
>>> my answer is that in the graph there is just one node pointed by two
>>> (or more) arcs (say =abc and =!1234), not two different nodes pointed
>>> by two different arcs. Just one node for two or more addresses (called
>>> synonyms), no redundancy.
>>>
>>> The issue, instead, is how to resolve two addresses to the same node.
>>> Semantically, resolution equivalence is given by owl:sameAs, so a
>>> browser knowing that one XRI address is the same as another will
>>> interpret the two XRIs as synonyms (in its internal data structure - a
>>> partial view of the global graph - the browser will build just one
>>> single node for the two addresses). If it does not know this, it will
>>> not assume they are synonyms and treat them differently - but it
>>> cannot assume they don't are unless explicitly stated (a principle
>>> called the "open world" assumption)
>>>
>>> Now, since XRIs are by definition abstract identifiers, to perform
>>> resolution they should be bound to a concrete resolution context. If
>>> WWW is the case, the resolution mechanism is through HTTP and the
>>> resolution context, I propose, could be based on Named Graphs (I'll
>>> come back on this with a concrete proposal in next weeks)
>>>
>>> I do not totally agree with you when you say:
>>>
>>>> The reasons in both cases is that the term "synonym" (and rdf:sameAs)
>>>> implied full logical equivalence -- essentially a merger of the graph
>>>> trees for both nodes. This is a beautiful concept, but very hard to
>>>> implement (as RDF folks have found out).
>>>
>>> clarified that what is described is "resolution equivalence", which
>>> is, as you describe, merging nodes from different trees to build the
>>> global graph, it is useful to point out that this mechanism is the
>>> very foundation of Linked Data: without merging there is no linking,
>>> so we cannot get rid of this. What we can do better than Linked Data
>>> (hopefully) is to clarify (at least some of) the many different usages
>>> of resolution equivalence (owl:sameAs) - as it has been pointed out in
>>> the paper by Halpin, Herman and Hayes
>>> (http://www.w3.org/2009/12/rdf-ws/papers/ws21) and internally
>>> discussed during many of our phcs.
>>>
>>> Best Regards,
>>> Giovanni
>>>