Error in RDF 1.1 Concepts and Abstract Syntax

Hi there,

I am writing at the suggestion of Ivan Herman regarding an error in the
definition of dataset isomorphism in
https://www.w3.org/TR/2014/REC-rdf11-concepts-20140225/#graph-isomorphism .

I include below an email correspondence that I have had with Ivan on August
28, 2019, including his proposed correction.

Best,
Rachel


From Ivan:
Rachel,

I think you are right, the definition in item (6) seems to be erronuous.

I believe the right formulation is something like

6. The named graph <n, G> is in D1 if an only if <M(n), M(G)> is D2

Which is all right, because a named graph is defined (a few paragraph
above) as "a pair consisting of an IRI or a blank node (the graph name),
and an RDF graph".

Your intuition is of course correct, but as this document does not define,
or refers to, quads, your definition would not hold either...

I would think that it is worth raising that erratum by sending a mail to
public-rdf-comments@w3.org, so that, eventually, it would find its place in
https://www.w3.org/2001/sw/wiki/RDF1.1_Errata. If, at some point, the RDF
spec is reopened by a WG, this should be taken care of.

Thanks!

Cheers

Ivan

On 27 Aug 2019, at 22:48, Rachel Arnold <rlongley@vt.edu> wrote:

Ivan,

Thanks again for pointing me toward the document with the standard
definition of dataset isomorphism. I have a question about the definition,
specifically statement 6.

There are a few things going on that are bothering me.

- n is never properly defined. Is it a node? Blank node?
- What does it mean for <n,G> to be *in* NG1? The opening statement of the
definition says that NG1 is a named graph. It seems that 6 is treating NG1
as if it is the collection of all named graphs.

Here’s what I *think* statement 6 is intending to communicate:

A triple t is in a named graph NG1 in D1 if and only if the triple M(t) is
in NG2 in D2 where M(NG1) = NG2.

Or perhaps more clearly:  <s,p,o,NG1> is in D1 if and only if
<M(s),p,M(o),M(NG1)> is in D2.

This makes me wonder - could statements 4 and 6 be replaced with

<s,p,o,NG1> is in D1 if and only if <M(s),p,M(o),M(NG1)> is in D2.

Or, maybe:

 There is a one to one correspondence between NG1 in D1 and NG2 in D2 such
that NG1 is RDF graph-isomorphic to NG2.

I’d like to get this pinned down before I spend too much time translating
my definitions into the standard, expected language.

Thanks in advance for the help.
Rachel

-- 

*Rachel F. Arnold, Ph.D.*

Collegiate Assistant Professor and GTA Coordinator
Department of Mathematics
540-231-8271 | 410 McBryde Hall



----
Ivan Herman, W3C
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
ORCID ID: https://orcid.org/0000-0003-0782-2704



-- 

*Rachel F. Arnold, Ph.D.*

Collegiate Assistant Professor and GTA Coordinator
Department of Mathematics
540-231-8271 | 410 McBryde Hall