Jack,
I will take the statement about uncountable identifiers with a grain of salt. I think “unbounded” is the proper mathematical term. But I contrast that hypothesis
with: Each topic in the topic map then provides mappings to all of the known UUIDs associated with it.
That “known” is a very strong limitation. We can start with “known to whom?” and then “how? what was compared?”
I will also disagree that a topic map can “handle ... every ontological definition”. The last time I looked, the expressiveness of topic maps was quite limited.
But, as someone else pointed out, ontologies tend to be either overspecified (technically precise) or underspecified (lots of primitive undefined terms). Topic
maps do not really support overspecification, and they encourage/demand some kind of definition. That is why they are ideal for Linked Open Data, which (IMO) does not really benefit from precise terms. You need enough precision to make critical distinctions,
but not so much that you reject viable candidate information sources. The discovery process for LOD involves kissing a few toads. The object is to maintain a profitable ratio of princes (or at least freeholders) to toads. YRMV.
-Ed
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx]
On Behalf Of Jack Park
Sent: Wednesday, October 22, 2014 2:58 PM
To: [ontolog-forum]
Subject: Re: [ontolog-forum] "Data/digital Object" Identities
It strikes me that metadata schemas are a great idea.
But, I would like to suggest that the number of ways in which individuals and groups identify the subjects in their universe might be uncountable.
My earlier comment spoke to the notion of using a topic map to handle not only every ontological definition, and every metadata schema, but also those "over-the-transom" representations that creep in over time.
The idea is to federate all ways of knowing without disturbing any particular identification scheme.
Each topic in the topic map then provides mappings to all of the known UUIDs associated with it.
That, in my view, supports discovery.
On Wed, Oct 22, 2014 at 11:34 AM, Gary Berg-Cross <gbergcross@xxxxxxxxx> wrote:
It is good to know of ITU-T's history of success. In this current proposal (ITU-T X.1255 (09/2013) 15) , as I understand it, they are discussing a "Framework for
discovery of identity management information."
They include some high level diagrams with semantic interoperability gestured at, but then provide nothing to advance on this front - it seems to me. Even there
discussion of metadata seems to beg off filling in this frame, so what you wind up with is a Recommended Framework that serves some non-semantic aspects of identifiers and leaves the hard problems for someone else at some other time and place. To quote from
the document:
"Identifiers are one important ingredient to achieving metadata interoperability. However, certain other aspects of metadata interoperability, including those involving human definition and context of descriptions, are outside the scope of this Recommendation"
Now we've had some O-Summits where we note the value of adding semantics to some standards and this may be another candidate. So I just worry as William indicated
earlier that we get this push to have a standard and then later see if it works...Tp quote from the document:
"The process of arriving at an agreed upon set of metadata schemas will become a collaborative process in which the interested parties will contribute their knowledge of the attributes which must be covered by the schemas; " Sure, it is a collaboration!
Below, BTW, is a copy of their discussion of metadata and their schemas from which above quote comes:
"transformation of the metadata records or in the search, and whether the metadata records are
transformed by the contributing registries or by the collecting registry, are all implementation
details. There may be significant performance consequences, but the basic design should allow for
implementation variations.
The approach taken in this Recommendation does not in and of itself solve the problem of search
and retrieval across heterogeneous information systems, but it does provide a common framework
in which different approaches can be used. Indeed, it is likely that there is no single solution to the
problem and that the optimized approaches may vary with community of practice and subject area.
7.6 Metadata schemas
A major objective of this Recommendation is to provide a basis for defining a set of "high-level"
metadata schemas to support the discovery of information on: a) identifiers used in various IdM systems; b) identity providers; c) relying parties; and d) trust frameworks and other IdM systems at all levels, including policies, procedures and underlying technical
infrastructure. The necessary elements of these metadata schemas will be driven by specific usage scenarios, but will have to be
extensible at both the element and the schema level in order to support growth and change in a
dynamic area.
The various entities involved in identity management can each define their own specific schemas and, as needed, map them into these high-level standardized metadata schemas to describe their
services, policies and procedures, and register these descriptions in one or more of a set of federated
registries. These registries would support discovery services across the registered entities.
While it is possible that a single metadata schema could be created to accommodate all aspects of
IdM technologies, relevant organizations and associated policies and procedures, it is proposed to
begin with a single schema for each type of entity involved. The process of arriving at an agreed
upon set of metadata schemas will become a collaborative process in which the interested parties
will contribute their knowledge of the attributes which must be covered by the schemas; the
evolving schemas can then be tested against various usage scenarios to see if they do indeed provide
the needed information to support the discovery processes, and could then be augmented, if
appropriate.
7.7 Metadata interoperability
Identifiers are one important ingredient to achieving metadata interoperability. However, certain other aspects of metadata interoperability, including those involving human definition and context of descriptions, are outside the scope of this Recommendation.
Other attributes specified in
metadata, such as those describing or enabling a particular configuration e.g., a specific connection
mode and aggregation approach fall within the scope of registry operation. For the purposes of
managing metadata entities across various registries, metadata interoperability will be facilitated if
the collaborating parties decide on common metadata schemas. Metadata, then, will be managed as homogeneous entities, with registries interpreting and processing them in a consistent fashion.
SOCoP Executive Secretary
On Wed, Oct 22, 2014 at 1:36 PM, Barkmeyer, Edward J <edward.barkmeyer@xxxxxxxx> wrote:
One must realize that the ITU-T people are quite capable of enforcing their framework in their own
industry – they are the nationalized and private telecommunications providers worldwide. Their standards make Korean cell phones work in the EU, and allow my cell-phone to access a land line in Japan. Of the bullets below, they have had (a) and (b) since
1982, using “ASN.1 Object Identifiers”, and they are still widely used in telecomm. And one of the first adopted URNs was urn:oid:<ASN.1 Object Identifier>, so that they could use them as URIs. And their software people are very disciplined, because when
they have a system failure, the consequences are dramatic, as a recent AT&T event demonstrated. So, protocol changes are disruptive and dangerous, and they don’t do pointless changes based on the bandwagon of the week.
Note how effective that standard has been outside their industry. (Well, the healthcare information
exchange people and a few others used it widely in the 1990s.)
In product labeling land, there is another system: UPN barcodes, and in the finance world, there
are international standard stock identifiers, and then there are DUNS and DIRAC codes for organizations, large and small. And in all of these cases, and in the ITU case, the use of these identifiers is de rigueur in the industry. These schemes were devised
and promulgated in these industries long before URIs, because they were needed for unambiguous transactions. And they all have a common property – there is a single organization that provides the identifiers, for a (not necessarily inexpensive) fee, because
they provide a service to businesses. (In the stocks case, and the ITU case, there are multiple appointed providers.) And in those cases, there is one reference registry, or at least one that can locate all the others. These identifiers have become successful
in part because there is guaranteed support: business serving business on a contractual basis.
Part of the problem with identifiers in the Internet of Things is that we have no idea who all the
providers will be, or what kind of support they will provide. Experience teaches us that most software producers have no concept of warranty, and a minimal understanding of service. And we already have competing technologies: IRIs, UUIDs, RFIDs, and all
the traditional ones. So, in addition to understanding the concept of identity, we need to understand the pragmatics of identifiers. There is not going to be one ring to rule them all. And everything that can go wrong with identifiers will.
So, we really need to address the issue that Jack Ring mentioned – the problem of “discovery” based
on characteristics. The gas station at the corner of High Street and Route 45, as long as it is still there, will provide some version of the expected service, even if the brand has changed, the owners have changed, the DUNSID has changed, and the city has
renumbered the addresses on High Street.
And OBTW, I think that topic maps might really be a useful and successful technology for “discovery”
in Linked Open Data, and perhaps in the Clouds. It is accessible to many more software engineers than useful OWL models, for example, would be. (And I am sure there must be existing projects, but I am too lazy to Google it.)
-Ed
Thanks all for the interesting comments on this which seems to agree that as William put it:
"identification in cyberspace is what creates the the acute need for better understanding of identifications."
I do find that computer scientists, programmers and the like often get off on a fast track without understanding some of the down side of ignored
issues.
They start talking about identifiers of digital objects and then jump to this as some basis for interoperability as in:
"The system of federated registries in this Recommendation is based on an open architecture that enables interoperability across arbitrary information systems."
It is in part a misplaced reliance on concretized identity management (identifier is a series of bits) as opposed to understanding that these identities (rather than The Identity) arise from a process of identifying something is some useful way among may such
ways as Hans noted.
BTW, if one wants to see some of that people are talking about as part of identity and ID management for digital objects you can look
at:
Since William has a long time interest following his Bachelor's work on some of this his comments may be interesting.
The above quote on interoperability is from this report as is this large claim (pg 7):
"The components are:
a) a scalable and distributed identifier system for the identification of DEs and for identifier resolution;
b) repositories for access to and management of digital entities; and
c) registries for federated search and discovery. Using these components, the resulting
distributed system can be managed through interface specifications and protocols instead of through the on-going maintenance of specific components.
Digital entities are the core element around which all other components and services are built and managed. Digital entities do not replace existing formats and data structures, but instead provide a common means for representing these formats and structures,
allowing them to be uniformly interpreted and thus moveable in and out of various heterogeneous information systems and across changes in systems over time."
SOCoP Executive Secretary
On Wed, Oct 22, 2014 at 9:47 AM, William Frank <williamf.frank@xxxxxxxxx> wrote:
Cyber Identity has been at the heart of my job for the last three years, and identity an interest ever since I wrote my bachelor's thesis on Leibniz and Master's thesis on Frege.
I agree with you, entirely Hans, and would say that implicit in your language is the essence of the problem: 'identity' is a kind of a very ghostly abstraction without much mooring.
Identity is surely not an attribute of a thing. What HAS a mooring is the ACT of identification As you put it Hans, "assigning an identity." The act of identification is, as you say, a social act, and is of course context dependent.
Also, I agree that identification in cyberspace is what creates the the acute need for better understanding of identifications.
But, it is not an entirely new problem, applications and deeper dives into what is already known might suffice.
For example, Gary's questions: Is a data object in one format the same as a data object in a different format or a different one? The bit streams can change but the original identity might be considered the same." This applies to *all* human artifacts. When
is Moby Dick the 'same' book? However, a new huge confusion has arisen, the conflating of
identifiers with identities.
As Gary says, 'seems like a large claim." Worse than large, if people think that computers can provide mathematical certainly about things in the real world, the assurance that, in effect, a passport MUST be a correct indentifier, then we are another step
along the way to handing over autocratic authority to the machines. Instead of 'we do not have a record of your payment'. we go do 'you did not make the payment.'
I am not sure how Jack's point about URIs relates, except that surely, 'to be is to be a URI" is another weird way the world might be going. For cyber thiings and their identifying
characteristics, I would agree with you, Jack. But, I also think that identification of *physical* objects might never or not for a long time be replicatable with information about the object that can be captured on a computer.
On Tue, Oct 21, 2014 at 7:22 PM, Jack Park <jackpark@xxxxxxxxx> wrote:
I've been importing ontologies into a topic map of late. It's rather surprising how many URIs have been assigned to the concept with the label "Person".
I think it is correct to argue that there are many different ways in which some entity is identified by different individuals and communities, so it would seem that any "Architecture"
which grows up around digital objects -- which, by many lights, are proxies for subjects one way or another-- should be capable of capturing all knowable ways to identify that object, regardless of the database identifier assigned to it locally.
Gary,
I tend to agree with your musings. The issue of identity (of whatever entity) is certainly one that
the network revolution has brought to increased importance, if for no other reason than that it exposes the identities that anyone assigns to an entity/object, be it digital or “real world” to those who may assign a different identity to that same entity/object,
however “sameness” might be defined. The NCOIC Net Centric Principles grappled with this issue by means of a principle called “Entity Primacy”, which basically states that whatever identity you might assign to an entity/object, it has other identities in other,
usually collective, frames of reference. Deal with that, as opposed to assuming that the identity you assigned has primacy. Usually that would mean recognizing that the entity/object has other identities in other frames of reference, and one should be prepared
to map the locally assigned identity to one or more other identities in other frames of reference, presumably those used by actors with whom one might want to exchange information about said entity/object.
Of course, one could argue that any entity/object has some “natural” or “inherent” identity, such
as the PID referenced below, UUID’s (Universal Unique Identifiers), or a person’s DNA, or perhaps more pragmatically, the VIN of an automobile. But even these assume a context of some, usually implicit, scope and an associated frame of reference. In other
words, such an identity is inherently one of the collective within which the entity/object is being identified. Entity Primacy therefore points out that no collective context has a priori primacy for assigning identities to entities/objects. One needs to specify
which collective context a particular identity for an entity/object is based on/derived from. And yes, this is recursive, since such collective contexts for assigning identities will themselves have identities in, presumably, larger contexts.
Humans just tend to glom onto some collective context (such as DNS) and assume that everyone else
will simply use that collective context for identifying entities, forgetting that not everything uses DNS, even in the networking domain. PIDs would certainly help things – but they are not universal and they likely assume some representational context dimensions,
as you surmise in your email. That’s OK as long as one is explicit about what those are and understand the scope limitations that they imply when interacting with others who might not share those assumptions.
Hans
From:
ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx]
On Behalf Of Gary Berg-Cross
Sent: Tuesday, October 21, 2014 5:39 PM
To: [ontolog-forum]
Subject: [ontolog-forum] "Data/digital Object" Identities
There is a bit of a movement to discuss digital data in terms of Digital Objects and an "Architecture." One rationale for this
seems to be to provide an easier mechanism for the "creation of, and access to, digital objects as discrete data structures with unique, resolvable identifiers" - From a CNRI’s Press Release.
It is further argued that such Digital Objects with a persistent ID (PID) will "provide a foundation for representing and interacting with information on the Internet."
Seems like a large claim and I wonder what this community thinks of this idea. After all Identity is quite a semantic issue and intuitions about identities for digital objects might
cause some problems. They seem quite mutable and we'd need to distinguish the ID for the raw data from each processing version of it. Is a data object in one format the same as a data object in a different format or a different one? The bit streams can
change but the original identity might be considered the same.
SOCoP Executive Secretary
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
|
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (01)
|