Re: Slash URIs and WebID Experiment

On 1/11/12 11:05 AM, Henry Story wrote:
> Why do you need to produce a 303 requiring URL? You can use a #tag url.
You are confusing matters.

In your world view, does a URL exist with a Fragment Identifier that 
works across the wire? In my world view, you are describing a URI that 
has mixed de-reference behavior comprised of:

1. across the wire address
2. local retrieved resource address.

The above works fine when you have a Web of Resources. The trouble is 
this: a Web of Linked Data is not a Web of Information Resources. It is 
a Web of Descriptor Resources where the Subjects of said Descriptor 
Resources are unambiguously Named using de-referencable URIs.

You have the following parts:

1. A Name
2. Address of a Descriptor (Information) Resource that describes a 
Subject by Name.

HTTP URI based Names allow a purpose specific role conflation that 
delivers the convenience of a network aware super key i.e., de-reference 
the key and you get the representation of its referent.

You can achieve the above, with HTTP scheme based URIs using:

1. # URIs - disambiguation of Name/Address is implicit due to the 
fragment identifier
2. Slash URIs - disambiguation is explicit via 303 based redirection.

Neither option is perfect.

For instance, if you want to make a Linked Data solution that isn't 
susceptible to the myriad of frameworks, libraries, browser versions 
(e.g., IE6) that handle fragment identifiers incorrectly (in the 
Information Space or Document Web context) then slash URIs are best. 
This is why DBpedia uses slash URIs. The goal was simple: it just had to 
work, period! Without that approach the LOD cloud [1] wouldn't have 
happened.

Downside of slash URIs, as you are seeing here, or will see eventually 
is this: the publisher has to implement the 303 heuristic for 
Name/Address disambiguation.

Above all, and @danbri has talked about this many a time in the past, 
Web users have already started using slash URIs as personal identifiers; 
exemplified by Web 2.0 patterns originating from the blogosphere.

One other thing, in the Web 2.0 realm, you have service home pages used 
by members as personal identifiers while inside those Web 2.0 systems 
they actually use mailto: URIs. Now, imagine a WebID spec that was 
cognizant of these realities, handled them deftly and gained nothing but 
mass adoption as its reward. Imagine that eh?

Just in case you don't understand why I am burning all this time on this 
matter, just remember the last sentence above. We (OpenLink Software) 
are committed to making that happen. I don't care if the end result is 
called WebID or NetID or something else.

-- 

Regards,

Kingsley Idehen	
Founder&  CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen