Re: SPARQL / Language spec ready for review

On Fri, Oct 08, 2004 at 05:37:17PM +0100, Seaborne, Andy wrote:
> 
> Simon,
> 
> Thanks for the comments so far.  Changes and non-changes described inline.
> 
> I'll carry over these comments when you send further reviewers feedback.
> 
> 	Andy
> 
> 
> Simon Raboczi wrote:
> >
> >On 02/10/2004, at 2:28, Seaborne, Andy wrote:
> >
> >
> >>The SPARQL language doc is ready for review in preparation for the 
> >>telcon next Tuesday.  Version v1.73 (or later) of:
> >>
> >>   http://www.w3.org/2001/sw/DataAccess/rq23/
> >
> >
> >These are my comments on revision 1.97.  I'm only up to section 2.2 so 
> >far, so I have to apologize that it's going to be later coming in than 
> >I estimated.  However, all my comments are on very minor issues.  I 
> >have yet to encounter an issue that would disincline me to endorse the 
> >document for publication.
> >
> >===
> >
> >[[ 1.97 -- Introduction
> >SPARQL is a query language for accessing such RDF graphs.  It provides 
> >facilities to:
> >  * select information
> >  * extract RDF subgraphs
> >  * construct new RDF graphs based on information from the target of the 
> >query.
> >]]
> >
> >"Select information" seems a trifle vague.
> 
> No change for now.  See below.
> 
> >
> >[[ 1.97 -- 2 Making Simple Queries
> >Patterns are descriptions of graphs with named variables in place of 
> >some of the graph labels or relationships; the simplest graph patterns 
> >are single triple patterns.  The graph labels are values as defined in 
> >[RDF Core concepts] and are URIs, plain literals and typed literals.  
> >Patterns can be combined using various operators into more complicated 
> >patterns.
> >]]
> >
> >I think it'd be more consistent to call these "graph patterns" rather 
> >than "patterns".
> 
> v1.103:
> Done.
> 
> >
> >This is my attempt to kill two birds with one stone, by pushing the 
> >definition of /graph label/ into 1 Introduction:
> 
> Graph labels considered harmful : see :
> 
> http://lists.w3.org/Archives/Public/public-rdf-dawg/2004OctDec/0054.html
> 
> This section has been updated based on Pat's comments.
> 
> >
> >[[ suggestion -- 1 Introduction
> >The subject, object and relationship components of graph statements may 
> >be labeled.  The /graph labels/ are values as defined in [RDF Core 
> >concepts] and are URIs, plain literals or typed literals.
> >
> >SPARQL is a query language for accessing such RDF graphs.  It provides 
> >facilities to:
> >  * select labeled components from a target graph
> >  * extract RDF subgraphs
> >  * construct new RDF graphs based on information from the target of the 
> >query.
> >]]
> >
> >[[ suggestion -- 2 Making Simple Queries
> >/Graph patterns/ are descriptions of graphs with named variables in 
> >place of some of the graph labels.  The simplest graph patterns are 
> >/triple patterns/ describing a single statement.  More complicated 
> >graph patterns are formed using various operators to combine triple 
> >patterns.
> >]]
> >
> >I use the term "statement" above because it was defined in 1 
> >Introduction, whereas "triple" hasn't been defined.
> 
> No change.
> Triple is from RDF Concepts.
> Use of statement has been eliminated from the document.
> 
> >
> >===
> >
> >2 Making Simple Queries
> >
> >The <div style="clear:both;"></div> appearing after the first set of 
> >example diagrams (triple1, triple2, triplePattern1) doesn't have the 
> >intended effect under either Safari 1.2.3 or Mozilla Firebird 0.7.  The 
> >first of the three diagrams is pushed up by the text "with result:".
> >
> >The resources <alice@work.example> and <robt@home.example> aren't valid 
> >URI references.  They should have a scheme part, presumably mailto:.
> 
> Can't change at the moment : I don't have the technology to change the 
> images.
> 
> For now, better to have the text and images agree even if both wrong, 
> rather than text right and images different.
1.06:

I updated the doc a bit to make sure that all the refs to the data in
that picture has "email addr" rather than <email addr>. Dan has
requested a starting query wihtout bNodes. Simon doesn't want us to
label the bNodes. As soon as I come up with the right graph and
pattern, I'll update the pictures, alt text and results. (Monkying
around with the pictures now will result in lost work.)

I say "right graph" because I'm looking for a graph where the graph
pattern is nore than one triple pattern, and will realistically match
more than one piece of data. I like foaf 'cause it's popular and we
can communicate more effectively if we talk about familiar data.

My currect favorite for the graph is still a foaf example where alice
knows bob and eve and the query doesn't concearn itself with labeling
the bNodes as the text never mentions them. Thoughts?

> >The labeling of blank nodes with _:1 and _:2 in both the triple 
> >diagrams and  the "who" column of the result gives the impression that 
> >the blank nodes have labels.  People desperately want to believe this, 
> >but it's not true and I'm uncomfortable about ever giving anyone an 
> >excuse to think so, doubly so in a W3C-sanctioned document.  Removing 
> >the _:1 and _:2 labels from the triple diagrams and leaving them in the 
> >result set would make me happier.  I prefer "harder to understand" over 
> >"easy to misunderstand".

Nack (for right now) but working on it.

> >I would prefer the diagram "query1" to be called "graphPattern1".
> 
> I'll leave this to Eric.
1.06:
done

> >It would be helpful to make some sort of explanation of the tabular 
> >result format the first time it occurs.  Something along the lines of 
> >"there is a column for every variable occurring in the graph pattern, a 
> >row for each solution; a graph label appearing in a column (variable) 
> >and row (solution) is the value bound to that variable in that 
> >solution".
> >
> >===
> >
> >2.1 Writing a Simple Query
> >
> >The data for the first example query should be specified to be in 
> >N-Triples format.  (It wouldn't hurt to similarly label the query as 
> >being actual SPARQL.)
> 
> No change for publication.
> I wanted to get into the example without lots of defining and padding text. 
> There is some explanation after the example.
> 
> >
> >===
> >
> >[[ 1.93 -- 2.1 Writing a Simple Query
> >... the '?' does not form part of the variables name.
> >]]
> >
> >"variables" should be "variable's"
> 
> v 1.103:
> Change made: "variables'" as the first use in the sentence is "Variables".
> 
> I see a panda :-)
> 
> >
> >===
> >
> >[[ 1.93 -- 2.1 Writing a Simple Query
> >The terms quoted by "<>" are....
> >...also possible to write floating point doubles directly.
> >]]
> >
> >I'd be inclined to pare this back to the absolute minimum to explain 
> >the preceding example, rather than trying to be a synopsis of the next 
> >several following paragraphs.
> 
> We have duplication between intro 2. and 2.1.  Not enough time to fix 
> before first WD publication.
> 
> >
> >[[ suggestion -- 2.1 Writing a Simple Query
> ><http://example.org/book/book1> and 
> ><http://purl.org/dc/elements/1.1/title> are RDF URI references.  
> >"SPARQL Tutorial" is a plain literal.  ?title is a named variable.
> >]]
> >
> >I'd then add a sentence about the format of URI references to the 
> >beginning of the paragraph about abbreviating them as QNames, and a 
> >paragraph describing the format of plain literals just before the one 
> >about typed literals.
> >
> >[[ suggestion -- 2.1 Writing a Simple Query
> >URI References are quoted within angled brackets in the same manner as 
> >in N-Triples.  Because URIRefs can become long, SPARQL provides an 
> >abbreviation mechanism....
> >]]
> 
> No change. <> for URIs is mentioned in RFC 2396 so it is much wider than 
> N-triples.
> 
> >
> >[[ suggestion -- 2.1 Writing a Simple Query
> >Plain literals are quoted within double quotes (") in the manner of 
> >N-Triples, using a backslash (\) to escape quotation marks and certain 
> >other characters.  As in N-Triples, plain literals may bear an optional 
> >language tag suffix introduced with '@'.
> >
> >RDF has typed literals.  Such literals are written using "^^"....
> >]]
> 
> There is (or will be) a reference to the formats mentioned.  Where they do 
> the "usual" thing, I'd like to keep this doc simple and not explicitly 
> mention it.
> 
> Also, "Plain literals are quoted within double quotes" needs to apply to 
> typed and tagged literals as well.
> 
> >
> >===
> >
> >[[ 1.93 -- 2.1 Writing a Simple Query
> >PRFEIX  : <http://example.org/book/>
> >]]
> >
> >"PRFEIX" should be "PREFIX"
> 
> v1.103:
> Done *2
> 
> >
> >===
> >
> >[[ 1.93 -- 2.1 Writing a Simple Query
> >Prefixes are syntactic...
> >... This query is equivalent to the previous one and will therefor have 
> >the same results:
> >
> >PREFIX  dcore:  <http://purl.org/dc/elements/1.1/>
> >PRFEIX xsd:       <http://www.w3.org/2001/XMLSchema#>
> >]]
> >
> >"therefor" should be "therefore"
> 
> ~v1.10: changed that text to:
> "will give the same results when applied to the same graph." to avoid "same 
> as" discussions.
> 
> 
> 
> >"PRFEIX" should be "PREFIX"
> 
> v1.103:
> Done
> 
> >
> >===
> >
> >(Continue with 2.2 and further tomorrow.)
> >
> >
> 

-- 
-eric

office: +81.466.49.1170 W3C, Keio Research Institute at SFC,
                        Shonan Fujisawa Campus, Keio University,
                        5322 Endo, Fujisawa, Kanagawa 252-8520
                        JAPAN
        +1.617.258.5741 NE43-344, MIT, Cambridge, MA 02144 USA
cell:   +1.857.222.5741 (does not work in Asia)

(eric@w3.org)
Feel free to forward this message to any list for any purpose other than
email address distribution.