On 11/4/10 10:22 AM, Ian Davis wrote:
On Thu, Nov 4, 2010 at 2:13 PM, Kingsley Idehen<[email protected]> wrote:
Ian,
Q: Is 303 really necessary?
A: Yes, it is.
Why? Read on...
I don't think you explain this in your email.
What's the problem with having many options re. mechanics for associating an
HTTP based Entity Name with a Descriptor Resource Address?
Do you mean associate a resource with a description? Or do you mean
something else? Can you rephrase using the terminology that everyone
else uses please.
Who is everyone else? How about the fact that terminology that you
presume to be common is actually uncommon across broader spectrum computing.
Anyway, translation:
What's the problem with having a variety of methods for using LINKs to
associate a "Non Information Resource" with an "Information Resource"
that describes it (i.e., carries its structured representation)? Why
place an implementation detail at the front of the Linked Data narrative?
We shouldn't be narrowing options for implementing the fundamental essence
of Linked Data -- hypermedia based data representation. Of course, we can
discuss and debate individual, product, or organization preferences etc..
But please lets not push these as mandates. We should never mandate that
303's are bad, never. Its an implementation detail, no more no less.
I'm suggesting that we relax a mandate to always use 303 and since
you're saying we must not narrow options then you seem to be
supporting my suggestion,
I didn't know there was a mandate to always use 303. Hence my comments.
The only thing that should be mandatory re. Linked Data is this: HTTP based
Entity Names should Resolve to structured Descriptors that are Human and/or
Machine decipherable.
Are you saying that requesting a URI should return a description document?
Resolve to a Descriptor Document which may exist in a variety of
formats. Likewise, Descriptor documents (RDF docs, for instance) should
clearly identify their Subject(s) via HTTP URI based Names.
Example (in this example we have 1:1 re. Entity Name and Descriptor for
sake of simplicity):
<http://dbpedia.org/resource/Paris> -- Name
<http://dbpedia.org/page/Paris> -- Descriptor Resource (HTML+RDFa) this
resource will expose other representations via <head/> (<link/> + @rel)
or "Link:" in response headers etc..
Ironically, bearing in mind my comments, we do arrive at the same
conclusion, but in different ways. I phrase my conclusion as: heuristics for
implementing HTTP based Entity Names that Resolve to structured Descriptor
Resources shouldn't dominate the Linked Data narrative, especially as
comprehension of the fundamental concept remains mercurial.
So are you contradicting your answer at the start of the post?
Huh?
I am saying, what I've already stated: heuristics re. essence of Linked
Data mechanics shouldn't front the conversation. You sort of arrive
there too, but we differ re. mandates.
Potential point of reconciliation:
You assumed that 303 is an existing mandate. I am totally unaware of any
such mandate.
I don't even buy into HTTP scheme based Names as a mandate, they simply
make the most sense courtesy of Web ubiquity. As is already the case
re., LINK semantics [1], Hammer stack [2], and WebFinger [3], mailto:
and acct: schemes work fine as Resolvable Names for Linked Data. In
addition, XRD [4] also works fine as Descriptor Doc option.
To sum the broader picture up here: let's be inclusive rather than
exclusive re. Linked Data.
Links:
1. http://tools.ietf.org/html/draft-nottingham-http-link-header-05
2. http://hueniverse.com/2009/11/the-discovery-protocol-stack-redux/
3. http://webfinger.org/
4. http://hueniverse.com/2009/03/the-discovery-protocol-stack/ .
--
Regards,
Kingsley Idehen
Ian
--
Regards,
Kingsley Idehen
President& CEO
OpenLink Software
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen
