Re: Contd: [pedantic-web] question about sioc / foaf usage

2009/12/1 Ian Davis <lists@iandavis.com>:
> On Mon, Nov 30, 2009 at 10:37 PM, Kingsley Idehen
> <kidehen@openlinksw.com> wrote:
>
>>
>> If you lookup Linked Data from spaces associated with myself of OpenLink you
>> will see use the aforementioned property re. missing relation. Also, you may
>> also find out that few people added the missing triple to their RDF files
>> after nudges from me.
>>
>> I hope I've made things clearer?
>
> I've read this thread and I don't understand the fuss. Some people
> aren't linking the document to the data it contains so we should
> encourage them to. Don't know why that is characterised as a debacle.
>

The necessary declaration of "document" as distinct, and yet necessary
for the definition of "data", and the necessity of different URI's for
these two concepts, are fundamental sticking points for many people.

If the HTTP web no longer existed (or the internet connection was
temporarily down), the discussion about document versus data would be
mute. Simple RDF Triple database queries, that do not rely on HTTP
communication, have no necessary need to refer to the
Document/Artifact. Only "data" would exist in the RDF triples (unless
you deliberately blur the division using the notion of foaf:Document
via foaf:primaryTopic for instance). Hence the debacle with saying
that Document is a necessary element to understand and use RDF data
linked together using resolvable HTTP URI's when to many it is just an
artifact that doesn't influence, and shouldn't need to semantically
interfere with, the data/information content that is actually being
referenced.

In the long term, I see it as introducing a permanent link from a
semantic RDF (or other similar format) universe to the current
document segregated web that wouldn't be there if everyone shared
their RDF information through some other system, and for example only
used the URI verbatim to do queries on some global hashtable/index
somewhere where there was no concept of document at the native RDF
level. The definition of Linked Data doesn't specifically say that
HTTP URI's have to be resolved using HTTP GET requests over TCP port
80 using DNS for an intermediate host name lookup as necessary, so why
should it require the notion of documents to be necessary containers
for data pretty much just because that is how HTTP GET semantics work.

I characterise it as a debacle because it has been a recurring
discussion for many years and shows that the semantic communicty
hasn't quite cleaned up its architecture/philosophy enough for it to
be clear to people who are trying to understand it and utilise it
without delving into philosophical debates.

Cheers,

Peter

Received on Tuesday, 1 December 2009 00:03:30 UTC