Re: Issue 42

I already said that both proposals (no-nulls & yes-nulls) can not be explained in terms of RDF semantics (see my three examples in the wiki). And I am NOT proposing to change SPARQL; the same way the proponent of the no-null DM are not proposing to change SPARQL. I don't know anymore how to explain that.
The only consequence I can draw from that is that we have to say that the DM does not deal with NULL values. 
--e.

On 6 Jun 2011, at 13:32, ashok malhotra wrote:

> Enrico:
> I see two problems with your proposals:
> a) They require a flag.  Earlier, in other contexts we had decided not to have flags (optionality) in the DM
> but wanted the DM to be a single, push-the-button algorithm.
> b) If we introduce a special constant, say, RDF:Null, we will have to explain this in the RDF semantics.
> We will also have to change SPARQL to take this new constant into account.  Unfortunately, we have
> no control over the RDF or the SPARQL WGs.  They are independent from us.  This is the more serious problem.
> All the best, Ashok
> 
> On 6/6/2011 3:01 AM, Enrico Franconi wrote:
>> Hi,
>> I do have in the wiki page three proposals.
>> Also, in the wiki there are my three examples still waiting for an answer by the people supporting the drop-the-NULL translation.
>> cheers
>> --e.
>> 
>> On 5 Jun 2011, at 09:52, Enrico Franconi wrote:
>> 
>>> On 5 Jun 2011, at 02:52, ashok malhotra wrote:
>>> 
>>>> I think a possible solution is:
>>>> 1) Do not generate any triples where the RDB has  NULLs
>>>> 2) Add a note that says, This rule is only one interpretation of NULL semantics.
>>> No. Unless there is an explanation on which "interpretation" are we talking about.
>>> At the current state of the (non)discussion, this "interpretation" is arbitrary, since it is unknown how to relate it to the semantics of the NULL of the rdb it comes from. So the note should say that it is unknown how to relate this interpretation to the semantics of the NULL of the rdb it comes from. It is important to warn the user about what he/she is buying with this interpretation.
>>> 
>>>> Users that prefer alternate NULL semantics should use R2RML.
>>> Before that:
>>> 1.5) as an alternative (via a flag) do generate triples where the RDB has NULLs using a special recognisable constant, with a note saying that in this case it is possible (by following some simple recipes for SPARQL or rules or other access methods) to relate this interpretation to the semantics of the NULL of the rdb it comes from.
>>> 
>>> cheers
>>> --e.
>> 
>

Received on Monday, 6 June 2011 13:28:46 UTC