Re: Alternatives to containers/collections (was Re: Requirements for a possible "RDF 2.0")

On Jan 16, 2010, at 2:07 AM, Dan Brickley wrote:

> On Fri, Jan 15, 2010 at 6:02 PM, Pat Hayes <phayes@ihmc.us> wrote:
>>
>> On Jan 15, 2010, at 3:36 AM, Danny Ayers wrote:
>> Well, OK, I guess there is a real split between the idea of RDF  
>> being simply
>> a datastructuring device along the lines of LISP or XML, and its  
>> being a
>> language with a semantics (and so supporting inference). Ive always  
>> assumed
>> that it was the latter. But if y'all think that the semantic web  
>> doesn't
>> need semantics after all, go ahead. I do have other things to do.
>
> It's a framework for describing stuff(*), the original '97 WG got that
> right in the naming, even if 'Resource description framework' isn't
> the most elegant of names. Once we agreed that for these purposes, all
> things are resources, at least that mission was somewhat clear.

Well OK, then it is a language with a semantics. Or it had better be,  
if what its doing is describing. Good, glad we got that cleared up.

>
> So what are the interesting characteristics of 'descriptions', in a
> computing context (setting aside poetry, rhetoric etc for now)?
> They're generally understood to be the kinds of thing that come from a
> point of view, can be rich or crappy, accurate or misleading, and are
> composed from smaller claims

That 'composed from' is exactly what the formal semantics specifies.  
Its often called a 'truth recursion': it says how the meanings of  
larger expressions is composed from those of the sub-expressions. For  
RDF it says that a triple means that a relation holds between two  
things, and a graph asserts all of its triples (ie, is true when they  
are all true.)

> whose accuracy and implications can be
> independently assessed, extracted or analysed.

Hmm. Why 'independently'? Sometimes, seems to me, the only way to  
assess the overall accuracy is to look at a whole graph of stuff all  
at the same time. This triple says that xxx is an :element, but what  
does *that* mean? Ah, over here, this triple says that :element is a  
kind of class containing :pthings, and lets see, what are those?...

> I'd try to keep these
> sorts of (very informal) observations in mind before going too far
> down the 'it's just a way to put data on the Web'. Zeros and ones are
> a way to put data on the Web. To have a framework for describing
> stuff, we need to have some associated practices for figuring out what
> those zeros and ones tell us about the world. When we're dealing with
> descriptions, ... they're descriptions 'of'/'about' something, and
> hopefully the acquisition of multiple descriptions should shed light
> either on the consistency or conflicts amongst those descriptions, or
> on the characteristics of the things they describe. Ideally both...

Well, the formal semantics can help with the consistency part, but Im  
not sure how to get a handle on the characteristic of the things  
described, except by supplying more descriptions. Maybe that would be  
a topic worth pushing on. RDF is just simple logical descriptions,  
right now, and its normative semantics would work just the same if it  
were all written down using oak-gall ink on vellum. But without the  
actual Web, RDF would hardly be the same, right? So what is it that  
the machinery of the Web provides, that makes a *semantic* difference?  
If Im writing a formal semantics, giving truth conditions for triples,  
what *more* should I be saying, that somehow takes advantage, or even  
just takes note, of the fact that this is all on an actual Web of HTTP  
traffic zooming all over the planet? For some things, this gives me  
actual access to the things described. Only for some, but OK, for  
those, what semantic difference does this accessibility make?

Pat

>
> Dan
>

------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Sunday, 17 January 2010 06:12:11 UTC