Re: Well Behaved RDF - Taming Blank Nodes, etc.

On Wed, 2012-12-12 at 22:37 -0800, Pat Hayes wrote:
[ . . . ]
> I also don't want any cycles, but that is much weaker than your
> proposal. Why not just say that wellbehaved means, no bnode cycles?

That would be fine from a formal perspective, and I first considered
proposing it that way.  The reasons I instead phrased the restriction in
terms of Turtle serialization were purely practical.  First, it seemed
easier to explain, to say that there simply cannot be any explicit blank
node labels.  If the restriction were expressed in terms of blank node
cycles, then there would have to be an explanation of what that means,
how the blank node graph relates to the original RDF graph, etc.  

Secondly, it is very easy to see, visually, whether a graph in Turtle
has any blank node labels.  In contrast, if the constraint were "no
cycles", then when you look at the Turtle the situation would be more
like: "Uh-oh, I see an explicit blank node.  Now I have to make sure
that it isn't causing a cycle."

Finally, in terms of promoting RDF uptake, IMO the less people have to
learn about, and deal with, the intricacies of blank nodes the better.
The mental gymnastics required to explain to someone how, as Nathan
Rixham aptly put it, a blank node "has a name, which isn't a name", are
in my opinion Not Worth It.  I can just picture some typical developer's
response: "WTF?!?!  Uh . . . thanks, but I have to get back to my XML
now."  Why should we expose users to such complications if we don't have
to?

In short, it is true that phrasing the restriction as "no explicit blank
nodes in Turtle" would prohibit some otherwise benign (acyclic) uses of
blank nodes -- and I'm open to phrasing the restriction differently --
but I think there is a lot of value in both providing very simple
guidance and (as much as possible) avoiding the whole issue of names
that aren't names.


-- 
David Booth, Ph.D.
http://dbooth.org/

Opinions expressed herein are those of the author and do not necessarily
reflect those of his employer.

Received on Thursday, 13 December 2012 17:10:28 UTC