Dan Brickley wrote:
On Sun, Apr 18, 2010 at 3:42 PM, Nathan <[email protected]> wrote:
Wonder what would happen if we just called them "Links"?
I think that would confuse people. And would put stress just on the
point where SemWeb and HTML notions of link diverge.
An HTML page can have two (hyper-)links, <a href="/contactus/">contact
us</a> in the header, and <a href="/contactus/">contacts</a> in the
footer. Each of those chunks of markup is what we informally call a
link; the relative URI reference inside the href attribute in both
cases is what makes it possible for the link to be useful. I'm saying
that http://example.com/contactus/ should be called a 'universal
resource linker' instead of 'uniform resource locator'. Using
'universal resource link' for that instead has a different grammatical
role and could confuse since the page has two links (the bits that go
blue in your browser usually), but they both point to the same
URI/URL.
Why do we need to change what a "URL" is? Education is a hard lengthy
and generally arduous process etc..
We simply need to build a bridge from where we are to where we are
headed using coherent terms etc..
A URL today enables you to locate a Generic Document (it may have a
myriad of purposes).
What comes next are Descriptor Documents accessible via URLs that have
the sole purpose of exposing a Structured Description of an Entity (Data
Object, Data Item, Datum).
URLs don't go away, they just provide us with access to "purpose
specific documents".
Seems to be pretty unambiguous, if I say "Link" to TimBL or my Mum they
both know what I mean, and it appears to produce the desired mental
picture when used.
There are two usages at least with link; 'pass me the link' versus
'click on the link'; the latter emphasises the occurance as being the
link.
We can "Refer To" and Entity or access its Structured Description via a
LINK when we construct a Descriptor Document using the EAV Data Model.
Link, short for HyperLink - Link as in Linked Data.
Keep the URI/URL/IRI for those who need to know the exact syntax of a Link.
So when the RDF perspective comes in, so do subtly different notions
of link. This is why I think framing 'link' as a countable thing will
lead to confusion. RDF links are a bit like relationships; so <a
href="http://bob.example.com/"; rel="xfn:coworker xfn:buddy">Bob</a> is
a link expressing two relationships, er, links. If you poke to hard at
the magic word "link" it kinda crumbles a bit. But it remains
incredible evocative and at the heart of both the Web and the SemWeb.
Linker is non-commital enough that allows a family of related
readings; where the markup describes a pre-existing link/relationship
(eg. co-worker), and where markup itself is the link we're interested
in.
If you check back to Timbl's original diagram in
http://www.w3.org/History/1989/proposal.html the different flavours of
'link' were in there from the start; 'wrote' and 'refers to' for
example; the former links a person to a document; the later connects
documents. So the linking story here is that identifiers for people
and documents can share a notation, and become linkable. What exactly
a link is, on the other hand, I think will always be a little bit
slippery.
Not slippery if we take this moment in time to apply some metaphorical
sand-paper :-)
Yes, <http://www.w3.org/History/1989/proposal.html> this document is
much clearer than the latest incarnation of the Linked Data meme IMHO.
Descriptor Documents are clearly visible in the early TimBL document,
for sure! The next step should have really been about, what constitutes
the Descriptor Document such that machines and humans can respectively
exploit these Descriptor Documents. Again, the answer
is/was/should-have-been: EAV Data Model.
Kingsley
cheers,
Dan
--
Regards,
Kingsley Idehen
President & CEO
OpenLink Software
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen
