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

Re: terminological clarification

From: pat hayes <phayes@ihmc.us>
Date: Thu, 24 Jul 2003 02:32:08 -0500
Message-Id: <p06001a0cbb4533aa8bf5@[10.0.100.23]>
To: Tim Bray <tbray@textuality.com>
Cc: www-tag@w3.org
>pat hayes wrote:
>
>>Let me illustrate the point with a simple example. If you click on 
>>http://www.ihmc.us/users/phayes/Yosemite.html
>>your web browser will show you a picture of Yosemite valley with 
>>some surrounding text which asserts, not unreasonably, that the web 
>>page you are looking at is about Yosemite. ...
>
>>Now, there are two ways we could use the above vocabulary to talk about this.
>
>I have a quibble with the presentation of one of the straw people, 
>but also I have questions.  Does the difference between these 
>stories matter?

Yes. For doing a formal semantics - which is a critical 
interoperability requirement for the SW, absolutely centrally 
important - it is vital to get this clear. I also think it is 
important for the architecture account more generally, actually: just 
to be clear about what the text is supposed to be saying.  Here's an 
example which I just found, in 
http://www.w3.org/2001/tag/doc/identify#sec-identify:

section 2 starts:
"On the web URIs identify resources."Any information that  can be 
named can be a resource." [RFC2396] ."
12 lines later it says:
"
In general resource is a time varying conceptual  mapping to a set of 
entities or values which are  equivalent [Fielding] ."

Wait a minute: something is wrong. *Any* information that can be 
named is a time-varying conceptual mapping...??? No, it's not: that 
is just obviously incorrect. There are all kinds of information that 
can be named that do not fit that description; the ancient pyramids 
are full of them, for example.  This kind of jarring conceptual 
mismatch keeps cropping up when trying to read this stuff.  If you 
read on, *all* the examples for the second definition are 
'information resources' which fit the first Yosemite story. There 
aren't any examples to justify the incredibly wide scope of the first 
claim.

>Does the difference between them have any observable effect on the 
>behavior of software?

Yes, absolutely.  Reasoning engines need to be sensitive to semantic 
constraints; sometimes even parsers need to be aware of them.  And it 
certainly has an effect on the behavior of people reading the specs.

>Most important: Is either of them a falsifiable hypothesis?

Well, I wasn't meaning to put them forward as empirically testable: I 
was trying to find out what you guys *mean*.  Clearly they are 
*different*, right?

But to answer directly, it seems to me that the first story is 
definitely falsifiable, and in fact the Web actually tests it every 
time anyone uses the URI to ping my server.  The second, referential, 
story is less empirically testable; but the point is that for the 
semantics to work, it - the claim that the web page is 'about 
Yosemite' - doesn't have to be provable; it just has to be 
*possible*. If URIs can only refer to one thing, and if that thing 
has to be an emitter of representations, the it can't possibly be 
about Yosemite; and that would be rather a semantic bummer.

>>First story (based on my understanding of REST). The "resource"  is 
>>an idealized abstraction of this page on my server, thought of as a 
>>kind of idealized Platonic document-in-the-sky (since this 
>>particular resource is static)
>
>Huh?  It's identified by a URI, it emits representations, that's all 
>there is.  The Platonic abstractions are your own invention.

Sorry, I didn't mean to sound flippant or sarcastic.  "Platonic" is 
just philosopher-talk for any abstraction: numbers are platonic.  I 
was trying to acknowledge the need to have an abstraction of the 
notion of a stored document which gets copied and sent to the client; 
the early, simple account that Roy discusses in his chapter 6 and 
which the REST model supplanted. I see that this notion of an 
'information resource' has to be described somewhat abstractly in 
order to cover things like webcams, continuously-updated information 
services, Google and the like, as well as plain old hypertext. And I 
was trying to say this without using the actual language of the REST 
model, because the whole point was that that actual *language* seems 
to be interpretable in more than one way.

But OK, "it emits representations" is enough to make it clearly not 
Yosemite valley, right? Which was the point.

Pat

>--
>Cheers, Tim Bray
>         (ongoing fragmented essay: http://www.tbray.org/ongoing/)


-- 
---------------------------------------------------------------------
IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes@ihmc.us       http://www.ihmc.us/users/phayes
Received on Thursday, 24 July 2003 03:32:12 UTC

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