RE: [xri] [Glossary] Definition of \"Resource\" and \"Attribute\"

[Drummond Reed <Drummond.Reed@onename.com>]

Xavier,

You're absolutely right. I've added your UML definitions of Aggregation and
Composition to the definition of Attribute. I think it helps harden it
nicely.

=Drummond 

-----Original Message-----
From: xavier serret [mailto:Xavier.SERRET@gemplus.com]
Sent: Thursday, March 06, 2003 4:02 AM
To: Wachob, Gabe; xri@lists.oasis-open.org; Drummond Reed
Subject: RE: [xri] [Glossary] Definition of \"Resource\" and \"Attribute\"

Yes,

This is in fact a very old topic, which I guess was first formally
introduced with database relational model.

Getting to something more recent, I would like to introduce the idea of
part and whole, which is UML way to address this issue.

We have to formal relationship specialization in UML to address this
relationship:
Aggregation & Composition:
Aggregation:
A form of association that represents a whole-part relationship between an
aggregate (the whole) and the constituent part.
Composition:
A form of aggregation with strong relationship and coincident lifetime
parts with the whole. A part may belong only to one composite.

The whole is an object that has "independent" existence.
Normally, you can always commute a composition relationship for a class
attribute. In UML this is normally just a design-time tradeoff.

So I would state following this rationale, that an attribute is a part of
a resource, which does not have existence by itself. 

So ... I would say that UML backs Drummond's proposal!

Xavier.

--
Xavier Serret.     Tel  +33 44236 4543
Corporate R&D Strategy @ Gemplus

-----Original Message-----
From: Wachob, Gabe [mailto:gwachob@visa.com]
Sent: Monday, March 03, 2003 8:03 PM
To: 'xri@lists.oasis-open.org'
Subject: RE: [xri] [Glossary] Definition of "Resource" and "Attribute"

I find it interesting that your definition has parallels in the Topic Maps
line of thought (where identifiers, associations, etc can be limited to a
certain "scope").

Anyway, at this point, I'm happy enough with your defintion - I think
perhaps we are at the level of nuances and can move forward.

        -Gabe

> -----Original Message-----
> From: Drummond Reed [mailto:Drummond.Reed@onename.com]
> Sent: Saturday, March 01, 2003 7:41 PM
> To: 'xri@lists.oasis-open.org'
> Subject: RE: [xri] [Glossary] Definition of "Resource" and "Attribute"
>
>
> Exactly right, Gabe, that's why I've been working to nail down the
> definition, because we have several important requirements
> pertaining to
> attributes.
>
> In fact, an offline conversation I had yesterday regarding attribute
> requirements illuminated for me precisely why it is
> important, and as a
> consequence suggested the most precise definition of
> "attribute" yet (at
> least within our context of identifiers and identification). To wit:
>
>       "An attribute is any data, metadata, or resource that can be
> identified only in the context of a specific resource.
> Attributes always
> express a relative relationship; they exist only in the context of the
> resource they describe."
>
> In other words, the issue is not whether attributes can be
> resources or
> resources can be attributes. They can. The defining quality
> of an attribute
> is the *scope of its identification*. What makes it an
> attribute is the fact
> that it can only be identified in the context of a specific resource.
>
> Example: John is a resource. Age is a resource. But John's
> age can only be
> identified in the context of John. Therefore John's age is an
> attribute of
> John.
>
> I hope this definition not only makes attributes clear, but
> also makes the
> requirement for attribute identifiers clear: it can be important
> syntactically to identity the scope of an attribute, i.e.,
> the resource to
> which it is relative. This would be easy if attributes were
> flat, but if
> they can be nested, then that syntactic delineation of the
> resource to which
> they are relative becomes even more important.
>
> Example: URI syntax delineates attributes from resources using the #
> character, i.e., the fragment delimiter. But URI syntax is
> ambiguous as the
> syntax of fragments, says nothing about them being nested,
> and is mute wrt
> versioning. These are all requirements I believe XRI syntax needs to
> address.
>
> =Drummond
>
> -----Original Message-----
> From: Wachob, Gabe [mailto:gwachob@visa.com]
> Sent: Friday, February 28, 2003 11:53 AM
> To: 'xri@lists.oasis-open.org'
> Subject: RE: [xri] [Glossary] Definition of "Resource" and "Attribute"
>
> Lets be specific about why we are talking about attributes:
>
> CR-6: Permanent Identifiers (resource stays a resource even
> if attributes
> change)
> PR-13: Attribute Identifiers (nested attributes? identifying
> attributes?)
> CR-14: Version Identification (identifying versions of attributes)
>
> Does this glossary definition addresss any ambiguity here? Does this
> definition clear up what "nested attributes" mean? What is an
> identified
> attribute? Isn't that a resource (any that has identity)?
>
> I'm not saying the definition you have doesn't work - I just
> want to make
> sure it addresses any ambiguity we see in the requirements
> we've actually
> identified.
>
>         -Gabe
>
> > -----Original Message-----
> > From: Drummond Reed [mailto:Drummond.Reed@onename.com]
> > Sent: Friday, February 28, 2003 12:16 AM
> > To: 'xri@lists.oasis-open.org'
> > Subject: RE: [xri] [Glossary] Definition of "Resource" and
> "Attribute"
> >
> >
> > Mike,
> >
> > Good point, use cases always help clarify terms. In fact
> > here's a good one
> > from the IETF ResCap spec
> > (http://www.ietf.org/internet-drafts/draft-ietf-rescap-req-01.
> > txt) that also
> > happens to include its own definition of attribute:
> >
> > "An attribute is a named characteristic of a resource
> > identifier with a
> > value that is meaningful in the context in which it is used.
> > An example of
> > an attribute might be a content type that an email address is
> > capable of
> > receiving."
> >
> > Note the phrase, "meaningful in the context in which it is
> > used". That means
> > the attribute "content type that an email address is capable
> > of receiving"
> > is an attribute of the resource "email address".
> >
> > So the use case is that a user agent such as a mail client
> > wants to retreive
> > an attribute - the supported content types - of a resource
> - an email
> > address. While the abstract concept "supported content types"
> > could be a
> > resource on its own, being able to resolve the abstract
> > concept "supported
> > content types" as a resource doesn't do the user agent any
> > good unless it
> > wants to know something about the schema of "supported
> > content types" - such
> > as the enumerations of MIME types.
> >
> > But in this case what the user agent needs is the supported
> > content types of
> > a specific email address - joe@example.com. This specific set
> > of supported
> > content types supported by joe@example.com can ONLY be
> > identified in the
> > context of that specific resource.
> >
> > So maybe I should modify my proposed definition of
> > "attribute" along the
> > lines of the one provided in ResCap:
> >
> >       "An attribute is an identifiable characteristic of a
> > resource whose
> > value is meaningful in the context of the resource it describes."
> >
> > =Drummond
> >
> >
> >
> > -----Original Message-----
> > From: Lindelsee, Mike [mailto:mlindels@visa.com]
> > Sent: Thursday, February 27, 2003 4:48 PM
> > To: 'xri@lists.oasis-open.org'
> > Subject: FW: [xri] [Glossary] Definition of "Resource" and
> "Attribute"
> >
> > Drummond,
> >
> > I understand the distinction you are making with the definition of
> > attribute, but I don't necessarily see the need for it.
> > Perhaps you could
> > walk me through a couple of use cases that would make clear
> > the need to
> > identify information exclusively in the context of a
> > resource.  I would also
> > like you to then show me how that wouldn't be already covered
> > by allowing
> > resources to "point" to other resources as "attributes."
> >
> > Mike
> >
> > > -----Original Message-----
> > > From: Drummond Reed [mailto:Drummond.Reed@onename.com]
> > > Sent: Thursday, February 27, 2003 3:59 PM
> > > To: 'xri@lists.oasis-open.org'
> > > Subject: RE: [xri] [Glossary] Definition of "Resource" and
> > "Attribute"
> > > (lo ng)
> > >
> > >
> > > Mike (and Gabe before him): good stuff. The main reason I
> > > sent that opening
> > > salvo was to help folks realize how large an impact to the
> > > whole effort even
> > > small issues with these core terms have.
> > >
> > > The second reason is that, while it might seem like the term
> > > "attribute" may
> > > not be that important, in my experience with both XNS and
> > > Liberty, it ends
> > > out being very important - almost as important as "resource".
> > > Being able to
> > > unambiguously and persistently reference an attribute in the
> > > context of a
> > > specific resource is critical when it comes to security,
> > > digital identity,
> > > DRM, and many other applications of XRIs.
> > >
> > > That said, I agree with you, Gabe, and Bernard that we should
> > > just stick
> > > with the URI spec definition of resource as "anything that
> > > has identity" and
> > > not try to define it further. It's not worth splitting hairs
> > > over whether
> > > simple attributes actually have identity outside of the
> > > resource that they
> > > describe.
> > >
> > > I think your definition of attribute as " data, metadata or
> > > other resources
> > > associated with a resource" is pretty close to the mark but
> > the words
> > > "associated with a resource" don't quite fully distinguish
> > > the two things I
> > > think are most important about attributes vs. resources:
> > >
> > > 1) Attributes are always relative, i.e., they only exist in
> > > the context of a
> > > specific resource, and
> > > 2) A special kind of attribute - an identifier - exists for
> > > the special
> > > purpose of forming an association with ANOTHER resource
> (that's our
> > > definition of identifier).
> > >
> > > To capture these two nuances, here's a modification to
> your proposed
> > > definition of "attribute":
> > >
> > >       Data, metadata or other resources that describe a
> > > specific resource.
> > > Attributes are always relative to the resource they describe.
> > > Identifiers
> > > are an attribute of one resource whose purpose is to form an
> > > association
> > > with another resource.
> > >
> > > How's that work?
> > >
> > > =Drummond
> > >
> > >
> >
>