Re: Fwd: Using Statement Identifiers to Manage Provenance

Olaf,

I think the issue of mutability/immutability is not primarily one of technical 
architecture, but one of management of resources.  (Like "cool URIs".)  I think 
that assuming a purely technical solution can ensure immutability on the Web 
will not work.  The best we might achieve with purely technical means is being 
able to detect when something has changed.

I think we should probably stand back a little from these debates on individual 
technical issues and start from a clear consensus of what we need to achieve, 
and how the web architectural framework shoes our options.  I'm planning a 
separate message on this later, but have another meeting right now.

#g
--


Olaf Hartig wrote:
> Hey Graham,
> 
> On Wednesday 18 May 2011 22:56:58 Graham Klyne wrote:
>> Olaf Hartig wrote:
>>> Hey Graham,
>>>
>>> On Wednesday 18 May 2011 11:01:17 Graham Klyne wrote:
>>>> [...]
>>>> My current sense is that RDF community consensus favours named graphs:
>>>> (1) the SPARQL syntax makes explicit provision for querying named
>>>> graphs, (2) the current RDF working group is giving consideration to
>>>> including a mechanism to encode named graphs within a single RDF
>>>> documemt
>>>> (3) even when using RDF without named graph support, named graphs map
>>>> directly to a natural web-based implementation: RDF documents
>>>> retrievable from web URIs.
>>>> [...]
>>> I never understood this argument completely so far.
>>> How exactly does this mapping you refer to work?
>> I assume you refer to (3) - it's simply if an RDF document is published on
>> the web using a URI, then that URI can be interpreted as denoting that
>> graph.  If that used in RDF statements published separately, those
>> statements can be metadata (e.g. provenance) about that graph.
> 
> And that's exactly where I have the following issue with this analogy: "an RDF 
> document is published on the web" is a Web resource, it's content (what we 
> will see a representations when we do an HTTP GET) may change over time. 
> Hence, you give a name (i.e. a URI) to something changeable. With Named 
> Graphs, in contrast, we name something which is immutable: a specific set (in 
> the mathematical sense) of RDF triples. This distinction may seem too subtle 
> but I say that it may make a significant difference when it comes to using the 
> name in statements about the named thing.
> 
> Olaf
>  
>> This is not always the most convenient approach.  I was just using it as an
>> illustration that named graphs are not a great distance away from plain RDF
>> on the web.
>>
>> #g
>> --
>

Received on Thursday, 19 May 2011 11:38:46 UTC