RE: About httpRange-14

Steve Pepper wrote:
> It's interesting to see this discussion surface yet again. It 
> would seem that the 303-solution has neither dispelled 
> anxieties nor quelled the debate.

Insightful comment. Sounds like you were not enamored with the solution back
when it was announced?

I'm curious; do you object to the context being determined by a dereference,
or something else?

> I would like to point out (once again [1]) that the cause of 
> the problem is the sleight-of-hand that took place when the 
> concept of resource was quietly redefined to mean "anything 
> that has identity" (i.e. anything whatsoever). This served to 
> blur the essential ontological distinction on the Web between
> (network-retrievable) information resources and "resources in 
> general".

Interesting article, thanks for sharing it. Frankly I didn't completely
follow it, I struggle with wrapping my head around RDF, but I hope I got the
gist of the article.

Question about [1]; who should be responsible to defines what an URI
identifies?  The owner of the URI, or the referer to a URI? If I read [1]
correctly it seems that you are proposing the referers are the ones that
should be defining identity, correct?  That seems wrong to me because of the
potential for many different referers.  How do you reconcile?

BTW, I had not heard of Topic Maps but read Wikipedia's entry and found
Ontopia's LTM[2].  It sounds a lot like RDF but for some reason I think I
could get my head around it about 5% more easily than I have been able to
with RDF.

> The HTTP URI is fundamentally an addressing mechanism, and it 
> is fine to use it as an identifier for the thing it 
> addresses. However, most of the things we want to address are 
> NOT network-retrievable and don't have an address. That 
> doesn't mean we can't use HTTP URIs to address arbitrary 
> subjects, but it does mean that the mechanism has to work 
> differently for direct and indirect identification.
> 
> That's why Topic Maps provides two URI-based mechanisms: 
> subject locators for direct addressing, and subject 
> identifiers for indirect addressing. RDF needs the same. The 
> 303-kludge (I'm sorry, but what else can I call it?) does not 
> work for humans, neither those that assign identifiers nor 
> those that need to interpret them; and it only works for 
> machines if you dereference every identifier before deciding 
> what to do with it, and that simply doesn't scale.

Okay, I think I can concur with you on RDF needed this.  

As for RDF, it seems there are two schools of though on RDF based on the
following extremes, and thus far I think I am in the latter; What of you?:

	1.) RDF is the foundation of all knowledge and we should quickly
convert everything possible to RDF. 
	2.) RDF is harmful and we should takes all efforts to existinguish
even its mention in history.

My issue with RDF is it seems far too complex to ever become mainstream
(compared to how mainstream HTML has become) and yet it seems to dominate a
lot of W3C standards development, likely because RDF is Sir Tim's baby.  Of
course publicly professing my concerns that RDF may in general be more
harmful for the Web than good is likely to ensure I'll never me on TimBL's
christmas card list.

However, both Topic Maps and RDF are a layer on top of the existing
HTTP+URL+HTML web, right?  I mean, we could easily have the web without RDF
and Topic Maps, right?  I would like to see a solution that can work with
just the knowledge the people have about HTTP+URL+HTML and not require them
to learn about triples or ontologies or directed graphs et. al.  Things that
are too complex for the average person just don't see mainstream uptake on
the web at large.

As I was writing this I started recognizing the inherent conflict between
those who need identities for URI to be precise (the RDF contingent), and
the rest of the human world where ambiguity can be resolved by context and
human intelligence in such a manner that almost nobody ever realizes that
there is any ambiguity at all.  But clearly given the state of things the
former have to be sated in order to address the needs of the latter, right?


A distinction that might be able to work for the RDF people and also permit
the type of uses that the rest of the world naturally gravitates toward
might be to introduce the concept of an "About" resource.  The URL for an
About resource would identify the document retrieved from that resource but
it could also be used as a proxy to identify the thing which it is "about."
For example, if I wanted to give you a URL to identify "me", I could give
you this URL as a proxy for me even though it identifies the document which
identifies me:

	http://www.mikeschinkel.com/about

How would this work?  It would be either based on 1.) context and/or 2.)
resolution, but ideally both.  As an example for #1 let's think about the
profiles at social networking sites where they ask people for their URLs.
People could start including their "About" URL in those fields.  And the
"About" URL distinction would be determined by either the person or entity
itself, or the owner or responsible body.  For me, I get to decide the URL.
For IBM, their Board of Directors get to decide.  For California, the state
government would decide, and they would also decide for Yellowstone Park.
For my car, I would decide at least until I sold it at which case it would
refer to my former car.

For #2, there could be an HTTP header or a new status code that says that
the resource is an "About" resource and that its URL can also be used to
refer to its document.

I believe having this concept of an "About" resource could empower humans
and organizations to start thinking about identifying themselves and the
things over which they hold domain, and then websites like Flickr and
YouTube and Facebook more could leverage these about URLs in a manner than
would be at least several orders more significant to the common man than the
concept of identify for RDF is today, and I think it could easily exist in a
manner that doesn't both RDF.  

Heck, RDF could officially choose not to recognize "About" resources and let
the web go about using them in a manner that average humans could comprehend
but still requiring URIs to be used for RDF as RDF requires. I'll bet if
there where such things as "About" URLs it is possible that the vision of
RDF and Topic Maps could eventually become mainstream after all by allowing
humans to get their heads about the concept of URLs identifying things and
having them do it all over the web in a manner that doesn't require people
to fully grok the RDF concepts (which I still don't.)  

The web standards elite could think of "About" resources as analogous to
HTML vs. XHTML.  The elite wanted people to write everything in XHTML
because it was so much "better" techinically yet (how many?) years later
people still write broken HTML because it doesn't require them to be so damn
precise, or because they just don't know any better and because companies
empower them to do so i.e. Microsoft's position on not letting bad markup
result in failure page view.  Similarly, people might start identifying
things with URLs if it didn't require them to think so much about it and
websites like social networks sites empowered them to do so. Heck, we'd
probably have to call them "About URLs" for the regular web user because
almost no one that is non-technical has any clue the meaning of "resource"
as it specifically relates to web.

Whew, sorry about that, I almost got off on a manifesto...  

There are a number of obvious issues I didn't address in my concept of
"About" resources, but rather than spend all the time on it I should first
see if people like the idea, or are strongly opposed to it and why. 

Thoughts?

That said, I also came across your writing [2] as I was finishing the above,
and in it you propose something you call a PSI ("Published Subject
Identifier") which sounds very similar to my "About" resource concept.
However, I couldn't tell if you were proposing a specific syntax for the PSI
or not, or if it was in fact very similar to an "About" resource?

-- 
-Mike Schinkel
http://www.mikeschinkel.com/blogs/
http://www.welldesignedurls.org
http://atlanta-web.org 

[1] http://www.ontopia.net/topicmaps/materials/identitycrisis.html 
[2] http://www.ontopia.net/download/ltm.html 
[3] http://www.ibiblio.org/hhalpin/irw2006/spepper.html
[4] http://thing-described-by.org/

P.S.  I also came across Thing-Defined-By.org [3] in your paper which gave
me a chuckle. It's an interesting concept and similar to Location.org
although Thing-Defined-By.org is meta and Location.org would be concrete. I
didn't particularly like some of his design decisions, i.e. to pass the full
URL scheme and all thus requiring encoding, but it was an interesting idea
nonetheless, but one that seemed very much inspired by the specific needs of
the RDF community as opposed to the general needs of web users who can deal
with ambiguity.

Received on Wednesday, 19 December 2007 12:09:14 UTC