Aaron,
Thanks for the term definitions. I was aware of them and was trying to
make clear the connection between them and similar thinking in other
domains. Sometimes I find myself, and colleagues in DL and Web world,
thinking we've invented ideas that have considerably more grounding and
historical development in other domains of discourse (dare I say even by
good old fashioned librarians!).
I agree with you that sadly namespaces seem now mostly to be tossed
around as weak syntactic sugar. No doubt you share my hope that this
will change and they might serve as some basis for recognizing
conceptual entities (as differentiated from their resolvable entities).
My argument is then what better vehicle for doing this than highly
recognized deployments like Dublin Core, OAI, and RSS (among others)?
Seems like we have a chance here to get it right and demonstrate to
other communities how that works. As I mentioned in my earlier note, we
in OAI turned to DCMI thinking they had it right but now realize that
they made a stab in the dark also.
If indeed, as you suggest, that "getting it right" is too much of a
burden for implementors then it seems that we might step back from our
semantic web visions and question the feasibility of its more mundane
but very important foundation technologies.
Carl
> -----Original Message-----
> From: Aaron Swartz [mailto:[log in to unmask]]
> Sent: Sunday, June 03, 2001 12:19 AM
> To: [log in to unmask]
> Subject: Re: DCMI namespace URIs
>
>
> First off, some clarification of terms:
>
> What you call concepts are traditionally known in the URI world as
> "resources" -- which are defined as "a conceptual mapping" which can
> represent anything that has "identity". What you call
> "manifestations" are
> traditionally called "entities".
>
> <[log in to unmask]> wrote:
>
> > Now for the lack of discipline. You suggest two
> namespaces: one for the
> > legacy Dublin Core (DC Classic?!) and one for other. The
> former seems
> > to be some historical artifact, the latter seems to be everything we
> > incrementally add as we "add new elements". Yet, if a namespace is
> > analogous to a concept space and a concept has some binding semantic
> > integrity (we hope), surely your criteria for placement in these
> > namespaces is troubling.
>
> I don't think so at all. I see the resource it represents as
> being rather
> close to "metadata concepts recommended at one point by the DCMI after
> 1999-07-02".
>
> I understand your search for cleanliness in clarity in the creation of
> namespaces. Sadly, as things currently stand, namespaces are
> a primarily
> syntactic notion, rather than a semantic one as you seem to
> want to make it.
> Think of this:
>
> For each new namespace we create, an RDF/XML developer has
> to add that
> namespace to their document to use those terms, plus keep
> track of which
> terms are in which namespace.
>
> For each namespace we change, we deprecate numerous
> existing documents and
> tools.
>
> I agree, the Dublin Core should have a strong semantic ground
> for its terms,
> their purpose, and their relationship to each other. However,
> I do not agree
> that this should carry over to its namespaces. To do so,
> would be too great
> a burden on implementers and users, in my opinion.
>
> > Finally, I'm confused about the concerns about changing our legacy
> > namespace URI. WOuldn't it possible to maintain the old
> namespace URI
> > (the one with the version # in it) and have it co-exist with a new
> > non-versioined namespace URI?
>
> I do not think the issue is having both namespaces co-exist,
> I think the
> issue is with why. If we did this, would we want people to go
> back and fix
> the many documents and tools that output this namespace?
> Worse, we have to
> fix the many tools (and educate the many people) that only
> understand the
> older namespace. Why all this pain for so little gain?
>
> > P.S. This same discussion is going on in OAI [3] where we
> are equally
> > confused about namespace URIs, decided to follow DCMIs precendent of
> > having version numbers in our URIs and now are reconsidering that.
>
> As you can probably see, proper use of namespaces is an
> evolving science for
> most people. ;-) My personal opinion is that you must change
> the namespace
> when you change the semantics of something that previously
> existed in it.
> Otherwise (when adding terms, for example) leave it the same.
> Oh, and be
> sure to document this process since not everyone has seen the
> light yet. ;-)
>
> --
> [ Aaron Swartz | [log in to unmask] | http://www.aaronsw.com ]
>
|