Re: Web Semantics for Datasets

On 2011-10-07, at 15:19, Sandro Hawke wrote:

> On Fri, 2011-10-07 at 09:12 -0400, Peter Frederick Patel-Schneider
> wrote:
>> For all HTTP clients?  Over all time?  For all
>> hardware/software/protocol/internet errors?  
>> 
>> Without knowing the boundaries of the "every" the proposal is
>> incomplete.
> 
> The goal here is to get systems to behave in a predictable (and useful)
> way so that other systems can build on them.  And the mechanism I'm
> trying to use for that is to define conformance to standards.  So, I'm
> thinking in terms of idealized behavior.    Certainly, it's out of scope
> to be worrying about whether, eg, client software has been subverted.
> 
> Can you think of a way to frame this better?   Mine was:
> 
>        For the graph naming to hold, every successful dereference of N
>        yielding an RDF graph must yield G. 
> 
> So, I'm not exactly defining all those terms, but I expect they are
> pretty clear.     Yes, for all HTTP clients.   Over all time -- that's a
> different kind of question, about the temporal range of the assertion.
> For all errors -- no, if there's an error it's not a "successful
> dereference".

There are some interesting failure cases there.

So, currently we use <$http_uri#date> as the graph identifier for our data generated from processing web pages. Right now, most of the time that will be fine, as that vast majority of those pages when dereferenced won't return any RDF data, but should RDFa become wildly popular, that would suddenly invalidate our choice of graph URIs. I'm not really happy that we could go from conforming to the spec/suggestion/design pattern to not conforming, because of external changes.

- Steve 

-- 
Steve Harris, CTO, Garlik Limited
1-3 Halford Road, Richmond, TW10 6AW, UK
+44 20 8439 8203  http://www.garlik.com/
Registered in England and Wales 535 7233 VAT # 849 0517 11
Registered office: Thames House, Portsmouth Road, Esher, Surrey, KT10 9AD

Received on Friday, 7 October 2011 14:33:11 UTC