Re: ACTION-133, ISSUE-22

Hi Eric,

I interpreted your text a bit and made changes to the draft. Relevant excerpts below. I'm setting ISSUE-22 to "Pending Review".

Best,
Richard


=== IN 4.2 SQL CONFORMANCE ===

This document specifies R2RML for databases that conform to Core SQL 2008 as defined in ISO/IEC 9075-1:2008 [SQL1] and ISO/IEC 9075-2:2008 [SQL2]. R2RML processors, R2RML mapping documents and R2RML mapping graphs may have to deviate from this specification in order to support databases that do not conform to this version of SQL.

Where SQL queries are embedded into R2RML mappings, SQL version identifiers can be used to indicate the specific version of SQL that is being used.


=== IN 5.2 QUERY-BASED TABLES ===

A SQL query-based table may have one or more SQL version identifiers. They must be valid IRIs and are represented as values of the rr:sqlVersion property. The following SQL version identifier indicates that the SQL query conforms to Core SQL 2008:

http://www.w3.org/ns/r2rml#SQL2008

The absence of a SQL version identifier indicates that no claim to Core SQL 2008 conformance is made.

Note: No further identifiers besides rr:SQL2008 are defined in this specification. The RDB2RDF Working Group intends to maintain a non-normative list of identifiers for other SQL versions [SQLIRIS].


=== IN INFORMATIVE REFERENCES ===

[SQLIRIS]
SQL Version IRIs, Members of the W3C RDB2RDF Working Group. The latest version is http://www.w3.org/2001/sw/rdb2rdf/wiki/SQL_Version_IRIs. This is a public wiki page.




On 21 Jun 2011, at 18:09, Eric Prud'hommeaux wrote:

> * Richard Cyganiak <richard@cyganiak.de> [2011-06-21 16:03+0100]
>> 
>> On 21 Jun 2011, at 15:43, Eric Prud'hommeaux wrote:
>>>> 3. It should say at least a sentence on where identifiers for other languages may come from.
>>> 
>>> I had deliberately left that information out because I didn't want to predict. Precedent indicates that sometimes these are strongly associated with a registry (as with the shortnames for XPointer registration schemes), defined by later revisions of standards (XHTML namespaces), invented by external organizations (as WS-1 did with SOAP-related identifiers) or invented by interested parties in the community (like most RDF identifiers or XML Signature's *Method identifiers).
>> 
>> It's quite a cop-out. As an implementer, what am I supposed to do if my implementation does special handling of a non-SQL-2008 dialect? There are some constraints (e.g., don't mint a new term in the w3.org/TR/r2rml namespace) as well as some guidance that we can give (e.g., look up some place to see if someone else minted an appropriate ID already).
>> 
>> Not saying anything is lazy and doesn't serve users or implementers well.
>> 
>> Can we operate a registry of SQL version IDs? Even if it's just a wiki page?
> 
> I think this is less dire than you express (many of the above extended protocols only defined a single identifier), but I'm happy to take a micro-formats wiki approach. Any reason this shouldn't be on the RDB2RDF wiki?
> 
> 
>>> Here's an example which relies on algebra syntax:
>>> _:subjMap1 rr:languageIdentifier <http://...relationalAlgebra> ;
>>>          rr:queryString "π ID,addr,city (People) ⋈ σ state="MA" Addresses" .
>>> 
>>> Critical? no.
>>> Useful? could be.
>> 
>> There are a lot of things that would be useful, but are out of scope. So is this one. If implementers want to use R2RML with relational algebra syntax, and have to do it this way:
>> 
>> _:subjMap1
>>  rr:sqlVersion <http://...relationalAlgebra> ;
>>  rr:sqlQuery "π ID,addr,city (People) ⋈ σ state="MA" Addresses" .
>> 
>> then yeah that's a mild form of abuse, but it *works* and that's good enough IMO, given that we don't have a requirement to actually support this kind of use.
> 
> I'm content to use the predicates rr:sqlVersion and rr:sqlQuery
> because it doesn't really break anything and it may be slightly easier
> for SQL users (the overwhelming majority of our clients) to understand
> the directives. Ivan, What do you think (noting that this is sort of a
> process question).
> 
> 
>> Oh and one more point: We need to settle wether languageIdentifier/sqlVersion can be multi-valued.
> 
> Yeah, I've been mulling that over in my head. I guess all of the use cases which motivate an identifier motivate n identifiers.
> 
>> Best,
>> Richard
> 
> -- 
> -ericP
>

Received on Tuesday, 5 July 2011 19:26:37 UTC