On 5/18/11 2:37 PM, Michael F Uschold wrote:
Alan, I'm glad you made that suggestion. I was also glad to see that
Tim-BL acknowledged that the URIs are just identifiers. As you know,
noone seems to be treating them that way, nor is there good tool
support to make it easy to do -- probably the main reason the practice
persists.
Here is the massive elephant in the Linked Data room:
do we continue to speak in the abstract about URIs or get down to the
nitty-gritty of dealing with its essential reality that's comprised of
the following:
1. Using links to craft whole data representation
2. Using links to cover functionality conventionally delivered by
de-reference (indirection) and address-of operators -- a feature native
to all programming languages, in some capacity .
Historically, the term URI is used loosely (i.e., with HTTP URI in
mind), and even worse, the tendency is to overlook (inadvertently) the
audience to which the Linked Data meme is being projected.
Micheal: your question basically highlights the problem I outline above.
The answer is dual pronged by the inherent nature of the URI abstraction.
Here I decompose:
Q: Should HTTP URIs be meaningful? A: Assuming 'meaningful' ==
understandable by the 'user' , what does this really mean? Remember, a
HTTP URI is endowed with Name/Address duality.
So we have a question answering a question due to the questions essence :-)
Q: Should URLs (a subClassOf URI) be meaningful? A: Assuming
'meaningful' == 'user' discernible and as a result hackable? Then the
answer is Yes!
Q: But aren't URLs just Identifiers? A: Yes, but they can be used in
special ways e.g. giving Names to Entities (what Linked Data is partly
about) as well as giving Names to Data Locations as per conventional Web
1.0/2.0 usage patterns.
I can go on....
Cut long story short.
URIs used as Entity Names should be treated as Identifiers in the purest
sense.
URIs as Entity Names in Linked Data Spaces have an implicit expectation
that de-reference (indirection) and address-of are combined such that
one Identifier is usable in the following operations:
1. de-reference (indirection)
2. address-of.
Due to HTTP ubiquity and implicit duality, an HTTP URI delivers the
above cost-effectively, but that isn't the only solution. Via custom
resolvers and alternative patterns e.g. .well-known/host-meta (a Web
Linking pattern) you can have Entity Names (beyond HTTP scheme) for
Linked Data Spaces that are discovered introspectively via linked data
resources (the actual data containers of EAV/SPO graphs in a variety of
formats). Basically, you discover the Name of an Entity Name via its
Data Representation Graph which is accessed via a URL (the Name of the
place from which you Access the Data Object).
Desperately hope this helps :-)
--
Regards,
Kingsley Idehen
President& CEO
OpenLink Software
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen
