Re: proposed change to a spec

On Mon, Feb 28, 2011 at 10:37 AM, Nathan <nathan@webr3.org> wrote:
> Jonathan Rees wrote:
>>
>> Nice try, Nathan, but I'm with David here. I think we have to assume
>> that a large amount of metadata has been deployed in good faith that
>> would only be true if representations were allowed to reflect
>> *partial* state.
>>
>> An example might be a blog. I'm not crazy about the practice of
>> confusing a blog with its home page but we may have to allow using the
>> home page URI to name the blog, with each representation carrying only
>> a portion of its state (state = its archive).  This is very similar to
>> the database use case in Roy's writings - database = resource,
>> representation = query page.
>>
>> Another example is using a table of contents with links instead of
>> putting the entire document on one page. Another is mobile phone
>> 'representations' of Wikipedia pages - those don't carry the entire
>> 'state' of the page.
>>
>> I reconcile these examples with a more principled view by imagining
>> that there must be *some* authorized representation that carries the
>> entire state; it just happens to be one that is never composed or
>> transmitted. (I knew that Aristophanes would serve me one day!)
>
> so then, we have to go to the inverse IR theory, which is that all
> information resources are Aristotelian abstractions,

I don't see why this is forced. You'll have to be more explicit if I'm
going to follow you.

For example, Ruttenberg's Scylla seems a perfectly fine model for IRs.
Maybe some IRs are Aristotelian abstractions, but other IRs might be
other things. You would have to argue that the Scylla cannot be the
right model, for some reason.

> this could fit if one
> said that the range of properties included the full set of representations
> over time, and all the properties of those representations (the hard cover,
> the draft, the paperback, the pdf, the kindle version, the html version on
> site x and the html sectioned one on site y).
>
> Okay, that brings me right back to my "as described by" train of thought
>
>  <u#h> is h as described by <u>
>  <u> is information
>  information serves to describe things

Stop right there.  Over my dead body.  You will not convince me that a
symphony or random word list or even a Lisp interpreter describes
anything.  That's a violent abuse of the word "describe".

*Some* information serves to describe things, sure.
http://odontomachus.wordpress.com/2011/02/21/the-place-of-metadata/

>  descriptions are a set of statements which serve as a (not the) description
> of something
>  descriptions can be comprised of statements
>  statements are declarations/remarks/presentations of opinion or position,
> offered as evidence that something is true
>  for statements to be offered means they must be made/said/offered by
> something
>  also means statements are open to interpretation (truth value established
> within a universe of interpretation)
>
> lol this is going to be fun.. so, the RDF <u> GET/200 OK resource
> representation case..
>
> the resource representation received is realization of a Platonic
> abstraction, namely a set of rdf statements. that set of statements forms a
> partial description of one or more things as they were perceived at some
> point in the past (reflects it's state), the set (or a subset) of those
> statements is open to interpretation and the truth value of those statements
> is determined within the universe of interpretation. Because the resource
> representation is but one resource representation in a potentially infinite
> set spanning both values and time, and because it's open to interpretation,
> the resource referred to by <u> is an Aristotelian abstraction.
>
> So, when Roy says to me "The clear understandable story is that resources
> are a continuum over time with myriad potential meanings depending on one's
> perspective." and "The Semantic Web, meanwhile, was created with no
> capability to express time, perspective, or change -- it is like trying to
> model life with nothing more than a few still pictures." - I've got to
> agree.

Roy is a great engineer, but I can only make sense of his ontological
statements ("resources are...") by reinterpreting them as software
design. I find most of what he says very difficult to apply to the
problems I care about, such as provenance and metadata
interoperability.

RDF is completely neutral when it comes to space, time, ontology, and
everything else. It is no better or worse at dealing with time than it
is at dealing with battleships.

You might argue that there are applications (maybe some involving
time) where you would want to use something other than FOL, or where
you might need 3-place predicates or product constructions or modes,
but that has nothing particularly to do with time.

A failure to get widespread adoption of any time-related vocabulary
and axioms might be due not to RDF but rather to some inherent
difficulty of the subject matter. Or perhaps there just hasn't been
the need, or no one has tried very hard.

> upshot of all of this, is that (1) information resources exist

I don't understand the reason for saying this, or what its
consequences are. Would it be OK if I mentally deleted it?

> and (2) the
> semantic web doesn't provision for the capability to express time,
> perspective or change. Which are all issues people have w/ RDF and that are
> being discussed at the minute, and which negatively impact linked data, such
> that the RDF semantics and concepts don't really provision for a read write
> web of linked data.
>
> now, ironically, that brings me right back to why I'm here, because I tried
> to use RDF for read write web of linked data and found I couldn't, and these
> issues are the ones that brought me to the point of talking to you guys
> about this.

OK.  Then we should talk about the application space directly, perhaps
as a separate thread since so far we've found the "no time" assumption
useful as a way to make an impossible problem almost tractable. I'm
not ready to jettison it yet for the main line of inquiry, but for an
application-driven second line it's a good thing to look at.

Remember that the main line, for me, is both much simpler and much
more important than time and PUT, so that is where I will put my
effort.

Some use cases would help a lot.

You know this, right?  http://www.ihmc.us/users/phayes/docs/timeCatalog.pdf

> conclusion: we need rdf 2.0 or "web data" with all these properties and tied
> in with the notion of authorative response, uri ownership and time.

umm... not sure where authoritative response and URI ownership come
from.  Don't you just mean nose-following?

My feeling is that "web data" can take care of itself but "data about
web" is a threatened species.

Best
Jonathan

> I'm happy with that, and it's been my gut instinct all along.
>
> Best,
>
> Nathan

Received on Monday, 28 February 2011 17:08:26 UTC