Re: Semantic web architectural requirement [was Re: Squaring the HTTP-range-14 circle]

Hi Alan,

On Wed, 2011-06-22 at 17:51 -0400, Alan Ruttenberg wrote:
> On Wed, Jun 22, 2011 at 2:45 PM, David Booth <david@dbooth.org> wrote:
> > On Wed, 2011-06-22 at 03:24 -0400, Alan Ruttenberg wrote:
> >> [ . . . ]
> >>  David wants us to believe that the writers of foaf,
> >> having omitted a disjointness axiom between documents and people, have
> >> deliberately sanctioned that they can be equated, solely based on that
> >> omission, and immune from analysis of any text or discussion that has
> >> been written on the matter.
> >
> > Well . . . not quite.  I'm not saying that the FOAF authors have
> > sanctioned this.  Indeed, they may well wish that RDF statement authors
> > would treat foaf:Document and foaf:Person as disjoint.
> 
> Then if someone treats them as the same, in a human-occupied forum,
> correct them instead of sanctioning their use. 

That is certainly a good thing to do when it is feasible, just as
agreeing on a common ontology is a good thing to do when it is feasible.
But as an architectural strategy it does not seem feasible to rely on
that approach when scaling up to achieve the full vision of the semantic
web: when it is a *machine* that is making use of many sources of data,
and you don't have the luxury of calling up a URI's human owner to ask
if they intended to omit that disjointness assertion.  This is the use
case that I meant.

> And don't call the result of their use "full fidelity".

Why not?  I made it very clear that I was using the term only in
reference to the *formally* expressed identity of the resource -- not
any additional intent that the URI owner may have.  This was clear both
in my original use of the term:
http://lists.w3.org/Archives/Public/public-awwsw/2011Jun/0006.html 
[[
 The client must be able to use a simple, architecturally
 authoritative algorithm to determine, with full [formal semantic] 
 fidelity, the URI owner's formally expressed identity for the resource.
]]

and in my detailed explanation of the phrase:
[[
What do I mean by "full fidelity"?  If both the publisher and the client
following the architecture and applicable standards then the client will
interpret the publisher's statements with the *same* formal semantics
that the publisher intended.  However, this does not -- and cannot --
extend beyond what is expressed in the machine-processable portion of
the statements.  It includes only what is expressed *formally* -- in
machine processable statements such as RDF or protocol codes.  It does
*not* include the human-oriented semantics of some natural language
prose embedded in an rdf:comment.  Note also that "full fidelity" does
*not* mean that the referent of a URI can be uniquely determined.
Rather, it means that its identity is constrained with the same
constraints -- neither more nor fewer.
]]

It is full fidelity in the same sense the digital communication is full
fidelity in comparison with analog: none of the digitally encode
information is lost, just as none of the *formally* stated information
is lost.  If you can think of a better term to use in this context then
please let me know. 

> 
> > I'm saying that RDF statement authors have no *obligation* to consider
> > anything beyond what is stated formally in the FOAF ontology.  I.e., as
> > long as the RDF statement author uses the FOAF URIs in a manner that is
> > consistent with the formal statements in the FOAF ontology and its
> > ontological closure (i.e., the transitive closure of the URI
> > declarations of all URIs that it uses), then the RDF author has met
> > his/her architectural obligation.  See Statement Author responsibilities
> > #3 and #4:
> > http://dbooth.org/2009/lifecycle/#resp3
> > http://dbooth.org/2009/lifecycle/#resp4
> 
> And I disagree. There is always the obligation to think. And where
> there are what seem to be inconsistencies between what is said in
> english and what is said in logic the responsibility should be to
> communicate with the author and work with them to improve the
> correspondence.

I've been thinking about this a bit, and it's not clear exactly what
you're disagreeing with.  It seems to me that there are two potential
obligations that we're discussing:

1. That the RDF statement author should use the URI in a manner that is
consistent with the formal assertions in the URI owner's definition of
that URI.

2. That the RDF statement author should use the URI in a manner that is
consistent with the informal intent of the URI's owner.

When #1 and #2 are not in conflict then there's no issue -- everybody's
happy. 

I agree that #2 is *nice* to do.  But I am saying that under normal
circumstances (i.e., except in cases of community expropriation as
described at
http://dbooth.org/2009/lifecycle/#expropriation ) #1 should take
precedence over #2.

Are you saying that you think #2 should take precedence over #1?  If so,
how would that be feasible if the URI owner's informal intent is
expressed in a foreign language that the RDF statement author does not
understand?  Or if the formal intent is scattered over thousands of
messages in an email archive?  Or if the RDF statement author cannot
find the URI owner's intent, or it isn't documented at all, and the
owner cannot be reached?   And how would that scale up to the full
vision of the semantic web, in which *machines* are doing this work?

> 
> > It would be *nice* if the RDF statement author read beyond the formal
> > statements to further divine the FOAF URI owners' intent, and the
> > statement author might be able to produce more useful data by doing so,
> > but the statement author has no *obligation* to do so.
> 
> As I have said previously, there is a social contract we are trying to
> establish, and this advise undermines it. In any case, the denial of
> obligation feels out of place. Obligation, afaik, doesn't play a role
> in any of the specification that we work with. Yet. If you want to
> propose this, then present it as a proposal, not as *the
> architecture*.

Okay, call it "social contract" instead of "obligation" if you prefer.
I'm not making a proposal any more than you are making a proposal for
what you think the social contract should be.  I am making observations
about what I see as *the architecture* for the vision of the semantic
web.  If you see it differently I'm very interested to know how.  And my
views may change over time.

> 
> > Again, the reasons for not requiring the statement author to read beyond
> > the formal statements are that: (a) reading beyond the formal statements
> > cannot be readily automated; and (b) different parties are likely to
> > interpret the URI owner's intent differently.
> 
> I have offered above what I consider to be good practice. I see your
> advise as "stick to the letter of the law, not the spirit".

To clarify my advice: Stick *first* to the letter of the law; if
feasible, also stick to the spirit.  


-- 
David Booth, Ph.D.
http://dbooth.org/

Opinions expressed herein are those of the author and do not necessarily
reflect those of his employer.

Received on Friday, 1 July 2011 17:58:46 UTC