W3C home > Mailing lists > Public > www-tag@w3.org > August 2003

Re: httpRange-14

From: Chris Lilley <chris@w3.org>
Date: Mon, 4 Aug 2003 15:51:26 +0200
Message-ID: <108265615625.20030804155126@w3.org>
To: Mark Baker <distobj@acm.org>
CC: www-tag@w3.org

On Friday, August 1, 2003, 11:03:33 PM, Mark wrote:

MB> I don't believe I've ever claimed that.  I claimed that
MB> http://www.markbaker.ca/ identified me, the person (as my FOAF
MB> asserts[1]).  I've also claimed that http://www.markbaker.ca/index.html
MB> identifies my web page, though I don't think I've mentioned that here.

MB> What you might have heard me say was something about how it's impossible
MB> to distinguish between me and my web page if the representations
MB> returned from GET on those two URIs have been equivalent, because
MB> its the representations that determine the "sameness".

This is where the definition of resource in the arch document (bits
plus metadata) helps, because those two representations are (slightly)
different - the bag o'bits is the same, most of the headers are the
same, but the one returned from http://www.markbaker.ca/ maybe also
has a Location: header or some other way to differentiate it?

Lets see

[clilley@tux clilley]$ telnet www.markbaker.ca 80
Trying 216.126.80.155...
Connected to markbaker.ca.
Escape character is '^]'.
GET / HTTP/1.0

HTTP/1.1 200 OK
Date: Mon, 04 Aug 2003 13:24:27 GMT
Server: Apache/1.3.23 (Unix)
Last-Modified: Fri, 01 Aug 2003 20:03:04 GMT
ETag: "1fc29-2a05-3f2ac778"
Accept-Ranges: bytes
Content-Length: 10757
Connection: close
Content-Type: text/html

<?xml version="1.0" encoding="iso-8859-1"?>
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<title>Mark Baker</title>
(and so on)

[clilley@tux clilley]$ telnet www.markbaker.ca 80
Trying 216.126.80.155...
Connected to markbaker.ca.
Escape character is '^]'.
GET /index.html HTTP/1.0

HTTP/1.1 200 OK
Date: Mon, 04 Aug 2003 13:26:22 GMT
Server: Apache/1.3.23 (Unix)
Last-Modified: Fri, 01 Aug 2003 20:03:04 GMT
ETag: "1fc29-2a05-3f2ac778"
Accept-Ranges: bytes
Content-Length: 10757
Connection: close
Content-Type: text/html

<?xml version="1.0" encoding="iso-8859-1"?>
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<title>Mark Baker</title>
(and so on)

OK so these are different URIs (strcmp) returning identical resources
(etags, LMD, etc).

Incidentally this seems related to siteData-36 in that it hilights
once again the impossibility of referring toa site, however that is
defined and most simply as a URI without a path, as distinct from the
welcome page of that site.

MB> Anyhow, this last exchange between you two has been particularly
MB> enlightening for me with respect to identifying what appear to be two
MB> major disconnects between you.  If you'll endulge me for a moment.

MB> Disconnect #1; Documents/Web-pages.  Culprit; Roy.

MB> As Tim mentioned, Roy does seem to be assuming that a timbl:web-page
MB> is what is returned from GET, a bag-o-bits.  Tim has said several
MB> times that it isn't, that's it's an abstraction, such as "a picture of
MB> Dan's car" in the abstract, rather than some particular PNG of it.

I think TimBL was applying that abstraction to 'resource' not
'representation'. One particular representation might well be
a PNG. What you get back from a GET is always a representation, not a
resource.

See diagram in
http://lists.w3.org/Archives/Public/www-tag/2003Aug/0018.html

BTW our archive system tries to put all image/* resources inline on the
img element and gives no other link to them or way to get at them
other then view source and hack the uri.

For the svg, see:
http://lists.w3.org/Archives/Public/www-tag/2003Aug/att-0018/uri-res-rep.svgz

MB> Disconnect #2; Identifying timbl:web-pages.  Culprit; Tim.

MB> Tim asked;
MB> "In your alternative architecture, where does one get the explict 
MB> information from - which allows one to refer to a web page by its URI 
MB> without meaning bridge or a person?"

MB> Roy has talked about this at least once on www-tag that I can
MB> remember (but can't find, sigh), but also in a recent discussion between
MB> the two of you, archived on www-archive.  He wrote;

MB>  "In any case, saying that "http" identifies an information resource
MB>   would not eliminate the indirection issue.  A document can talk about
MB>   some other document just as easily as a car.  We eliminate the 
MB>   indirection
MB>   case by declaring that assertions that target a URI are assertions on
MB>   the resource identified by that URI: state that is only reflected by the
MB>   content of all its representation over all time.  The only way to make
MB>   assertions about the information content returned by an action is to
MB>   add qualifiers for method and time, since the architecture requires
MB>   that those be orthogonal to the identifier."
MB>   -- http://lists.w3.org/Archives/Public/www-archive/2003Jul/0007.html

An assertion I would substantially agree with.

An interesting result of Roy's argument is that resources, since they
cannot be observed directly, can only be inferred through observation
of the representations that they have returned over time.

I think that is a valuable observation and leads further to a
principle that I find clarifying, though it may upset some: resources
are inherently second class objects. URIs are first class objects, and
may be stored, transferred, compared for equality and so forth. So can
resource representations (and proxy caches do this all the time).

Resources, though, are a derived abstraction. You can't observe them
directly, or measure them, or compare them. You can compare their
identifiers, but not the resources themselves. Their meaning, the
abstraction that they convey, is reverse engineered from observation
or from shared social discourse. The meaning is not observable from
the identifier in isolation, only from the usage of this identifier.

Naturally, social usage of an identifier is influenced by the
representations returned from it.

Consistency, then, is a social effect where observations on the usage
of the identifier and observations on the returned resources are
substantially in agreement. Consistency is an analogue quantity, not a
binary one - consistency can be greater or lesser, can be argued
about, has shades of meaning, and whether the consistency is good
enough depends on the use that will be made of it.

To oxaca weather example demonstrates this. The resource is never
actually stated or fully defined, merely inferred. "The weather in
oxaca" is not suitable as a definition, merely as an aid to social
discourse - an *environment of use* of the identifier. More precise
definitions (the weather at the weather station in oxaca, lattitude
this, longitude that, with temperatures updated every 20 minutes and
rainfall updated every 24 hours), or merely different definitions (a
prime example of the centralist tendency to asset that the center of a
town and it's suburbs have the same weather), can always be made.

MB> Hopefully that helps.  Sorry for butting in, but I'm just itching to
MB> see this resolved.

Please do butt in. Discussion helps.


-- 
 Chris                            mailto:chris@w3.org
Received on Monday, 4 August 2003 10:20:28 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:00 UTC