Re: solution modifiers & construct/describe

On Mon, 2005-06-27 at 15:04 -0400, Kendall Clark wrote:
> On Mon, Jun 27, 2005 at 07:54:54PM +0100, Seaborne, Andy wrote:
> 
> > SELECT and CONSTRUCT aren't so very different:
> 
> Hmm, really?

I had similar concerns when I reviewed the definitions a while
ago. I sent a big pile of suggestions; Andy revised the definitions
somewhat like what I suggested and somewhat differently. I haven't
gone back over the whole thing to check the details, but I think
the definitions give a reasonable account of the interaction
between solution modifiers and construct/describe.

> I think at the very least the interaction of the SMs with DESCRIBE &
> CONSTRUCT needs to be spelled out very explicitly in the spec.

Yes, I think another 2 or 3 or 4 examples would be worthwhile,
since both Kendall and I found this aspect of the spec a bit
puzzling.


>  I further
> think some of the SMs should be restricted w/r/t some query forms.

As to whether solution modifiers *should* be allowed on
construct/describe, I don't have a strong opinion, though
I have a vague memory that Jos argued in favor at one point.

(well, by way of full disclosure, to restrict the query forms
is a design change that will undo the LC decision we made a little
while back, so I'm disinclined to go there, on scheduling grounds.
But that argument shouldn't weigh heavily.)

>  Does CONSTRUCT really return an RDF graph? I simply don't
> understand the point of saying that you want an RDF graph ordered by
> descending ?hits order. There's no sense, as I understand these things, of
> ordering triples in an RDF graph.
> 
> (I can see the point of limit, however.)

It's not that the graph is ordered, but ORDER BY interacts with LIMIT
to say which 10 ?hits you're going to get in the graph.


> And DESCRIBE seems even worse, actually. It's *so* unconstrained
> semantically in the spec (by design, of course) that it has the problems of
> construct re: ordering, but has other problems, IMO, re: limit. If you have
> no idea what you will get back from DESCRIBE, what sense does it make to say
> you want those unknown results ordered, limited, or offset?

I'm already on record objecting to the whole notion of DESCRIBE, so I'll
leave that be.


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Monday, 27 June 2005 19:32:27 UTC