OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

# xdi message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]

Subject: Re: [xdi] EquivalenceLinks - XDI Wiki

• To: Drummond Reed <drummond.reed@xdi.org>
• Date: Sat, 4 May 2013 23:02:55 +0200

I think it's okay to say that =a/\$ref/=b and =b/\$ref/=c does imply =a/\$ref/=c.

But it doesn't mean there's actually a =a/\$ref/=c arc in the graph.

Markus

On Sat, May 4, 2013 at 10:38 PM, Drummond Reed wrote:
On Sat, May 4, 2013 at 1:23 PM, Joseph Boyle wrote:
Yes, I think this is clearer than link. Although in math http://en.wikipedia.org/wiki/Equivalence_relation implies reflexive, symmetric, and transitive. \$is meets all three conditions while \$ref and \$rep meet none of them.

Part of the reason we created the https://wiki.oasis-open.org/xdi/EquivalenceRelations page is to define the different types of equivalence that XDI supports. We split off first \$ref and then \$rep so that \$is could meet the very strict reflexive, symmetric, and transitive requirements of a "Leibniz equivalence" test. \$ref and \$rep are not supposed to meet these requirements.

Actually, when we say \$ref is non-transitive, we are saying =a/\$ref/=b and =b/\$ref/=c does not imply =a/\$ref/=c. However is it even possible to have both =a/\$ref/=b and =b/\$ref/=c in the first place? If not, how are we avoiding it? If it is possible, then how do they behave?

Actually, reading this, I think we were wrong to say \$ref is non-transitive, i.e., =a/\$ref/=b and =b/\$ref/=c does imply =a/\$ref/=c. Does anybody disagree?

If not, Joseph, can you update that page to correct this?

Thanks,

=Drummond

On May 4, 2013, at 1:02 PM, Drummond Reed <drummond.reed@xdi.org> wrote:

I believe it's important to keep XDI literal nodes fully aligned with RDF literal nodes.

\$ref and \$rep (and \$is) are, IMHO, not special types of nodes, but special types of relations. That's why we have special treatment for them.

In fact, this is such a good point that I don't know why we originally decided to call them "equivalence links". To be consistent with the XDI graph model should have called them "equivalence relations".

So I just did a search-and-replace edit on the page to change "links" to "relations" (it reads wonderfully) and then renamed the page to:

Just FYI, to rename a page on the wiki requires two steps:
1. Use the "Copy Page" function on the drop-down menu to copy the page to the new page name.
2. Edit the old page to replace all the content with a single line:
#REDIRECT NewPageName

=Drummond

On Sat, May 4, 2013 at 2:31 AM, Joseph Boyle wrote:
Drummond reminded me that \$ref or \$rep preclude not only literal but any other child nodes. This is such a strong constraint that a node originating a \$ref or \$rep looks to me less like other context nodes, and if anything more like literal nodes. Can we make them an additional category of nodes, on a par with context nodes and literal nodes? Better yet, can we subsume literal nodes and alias nodes (originating a \$ref or \$rep) under a category called value nodes.

On May 3, 2013, at 3:56 PM, Drummond Reed <drummond.reed@xdi.org> wrote:

Markus, I was just discussing this with Joseph and I think he's got a point that, in that proposal, we'd be better off calling them equivalence arcs than equivalence links. All arcs are a form of link, but "link" has a lot of Web semantics, whereas "arc" is pretty much just a standard graph concept.

What do you think?

=Drummond

On Fri, May 3, 2013 at 7:48 AM, Markus Sabadello wrote:
I would say "arc" and more specifically "relational arc" are low level terms in the graph model. We can always use these terms. An "equivalence link" is a relational arc that expresses equivalence. I don't think we're using just the term "link" by itself.

Markus

On Friday, May 3, 2013, Joseph Boyle <planetwork@josephboyle.net> wrote:
> When should we say "link" vs. saying "arc" or "relational arc"?
>
>
>
> Sent from my iPhone
> ---------------------------------------------------------------------
> To unsubscribe from this mail list, you must leave the OASIS TC that
> https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
>
>

--
Project Danube: http://projectdanube.org
Personal Data Ecosystem Consortium: http://personaldataecosystem.org/

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]