RE: Arch Doc: 26 September 2003 Editor's Draft (review of someterms)

On Wed, 2003-10-01 at 06:34, Olivier Fehr wrote:
> Ian,
> 
> -----Original Message-----
> From: Ian B. Jacobs [mailto:ij@w3.org] 
> Sent: mardi, 30. septembre 2003 23:51
> To: Olivier Fehr
> Cc: Dan Connolly; www-tag@w3.org
> 
> On Mon, 2003-09-29 at 17:02, Olivier Fehr wrote:
> > Abstract
> > <quote>
> > ...of Web resources that are interconnected via URIs..
> > </quote>
> > ->probably better to say 'can be interconnected', as they may exist in
> > the same information space without any relation to each other.
> 
> 
> I propose to change "that are" to ", which are".
> [Olivier says] Not sure what the underlying concept is that makes you
> say that Web resources _are_ interconnected. In what sense?

Interconnected via URI references. I must be missing what question
you are getting at.

> > 2.1 Comparing Identifiers
> > <quote>
> > ..,it is generally not possible to be sure that two URIs that are not
> > equivalent identify different resources.
> > </quote>
> > This follows from 'Web architecture does not constrain resources...'
> in
> > 2.
> 
> Does it? What if the Web architecture allowed us to use more than
> one URI to identify a resource, but in all cases it was possible
> to determine that from examining the URIs alone. I don't think
> the part about "generally not possible to be sure" is necessarily
> implied by the sentence you are referring to.
> [Olivier says] To be more specific, if I accept that I do not have any
> authority over www.w3.org, then I cannot rely on any assumption I make
> about www.w3.org/whatever unless the authority describes somehow what
> /whatever is, but then I no longer have to make any assumption.
> My point is, if there is no central authority that can determine what
> www.w3.org/whatever is supposed to be and how it is supposed to look
> like then we must rely on the 'decentralized' authorities.
> If I understand, your approach would be somehow defining/mandating the
> meaning of /whatever centrally? 

Only in that using the Web means agreeing to use (certain)
specifications. I'm not sure what part of the text implies
(in the general case) that a central authority determines
whether two URIs identify the same resource.

> > <quote>
> > Agents should not assume...
> > /Oaxaca and
> > /Oaxaca identify the same resource,...
> > Thus, the parties responsible for weather.example.com should not
> use...
> > </quote>
> > I don't think the 'thus' is necessarily correct, as the responsible
> for
> > weather.example.com can simple determine, i.e. stipulate that both
> > reference the same, whereas a agent is not free to assume that
> > equivalence lacking a clear statement of equivalence from the
> > responsible authority.
> 
> I'm not sure I've understood your point, since you seem to
> be agreeing with the point of the paragraph.[Olivier says] 
> 
> [Olivier says] I am getting at is that the agent is not free to assume
> the meaning of, what is behind /whatever, even if it seems obvious (this
> is seen from the view of the agent). 

> From the view of the authority, why
> should the authority not be able to declare, for example, /Whatever and
> /Whatever equal. So I can't see why ...the responsible for ... should
> not use... 

The authority can declare that uRi and UrI identify the same
resource. But when there's no way for an agent to determine 
the equivalence relationship by examining the URI alone in ways 
licensed by specifications, then the authority has made life
difficult for the agent. So we recommend against publication
of arbitrarily different URIs for the same resource. It seems
to impose a cost and offer no benefits.

> > 2.2 URI Opacity
> > 'Good practice'
> > Somehow confirms my believe above
> > <quote>
> > The example URI used in ... suggests...
> > On the other hand, the "mailto"...
> > </quote>
> > As you say, the normative specification makes the difference. By
> > assuming that a certain type of resource can be constructed by
> guessing
> > the parts of an URI, an agent might well end up with something
> > completely irrelevant to her/him/it.
> 
> I'm not sure if you are proposing a specific change here.
> [Olivier says] The agent should (be able to) rely on the authority's
> definition what a resource under its authority is about. 

> Obviously, they
> will need to have that description to be available in a standard format
> adhering to a defined vocabulary, however if you standardize this
> vocabulary, without possibility for extension, then authority is no
> longer 'in control'. 
> In short, is the goal to mandate via a standard a complete vocabulary or
> a just rules how to construct one? In the later case the authority would
> probably make changes which the agent may not be able to deal with. I
> gues we would have to live with this coupling, unless the agent is able
> to adjust itself to new vabularies 'on-the-fly', e.g. by interpreting
> the authorities description according to some rules that may possible be
> supplied with the description. Makes the agent a behave more like a
> human, I think...

Norm Walsh and Dave Orchard are developing a TAG finding on these
topics. I look forward to addressing some of these questions in the 
Arch Doc but also pushing some discussion out to the finding.

 _ Ian
-- 
Ian Jacobs (ij@w3.org)   http://www.w3.org/People/Jacobs
Tel:                     +1 718 260-9447