Re: in defense of standards

On Friday, October 10, 2003, at 09:41  AM, John Black wrote:

> From: Bijan Parsia [mailto:bparsia@isr.umd.edu]
>> Sent: Thursday, October 09, 2003 10:43 PM
>> To: Sandro Hawke
>>
>> On Thursday, October 9, 2003, at 10:18 PM, Sandro Hawke wrote:
>>
>>> and you're still going to build an open systems, I think you're
>>> just arguing for an undocumented protocol.
>>
>> Or evolved ones. Or these aren't "protocols" in the networky sense.
>
> Ok.  I'm willing to join the bands of the freedom-loving-formalists,

well, we may be freedom-loving, but we are exclusionary elitists as 
well.

> but I have some questions.  Please forgive me here, I am clearly the
> least experienced person on this list, but I don't know where you are
> getting this dispute protocol from.

What protocol?

We're just pointing out that closing things down a certain way 
precludes other desirable choices. None of us claim the Semantic Web 
"stack" is done.

> I don't remember reading anything
> in the proposed standards about rdf:dispute or owl:disagreement.  Is
> there such a thing?

No. There doesn't need to be, of course. It might not be reflected in 
the formalism. There's no owl:consistent, or rdf:entailment.

Think of all the things we'd have to have in the language to express 
(or even use) such things. Contexts at the *very* least.

>  I would like to be able to use your dispute protocol
> and to program my agents to use it.

I can't tell if you're funning me or setting up a reductio, or what.

But ok, here's one: If you see that someone uses a URI, and import a 
document that provides an ontology that, when imported into the "home 
ontology" of your URI is inconsisten, where either ontology alone is 
consistent, then you have a disagreement between the user of the URI 
and the owner of the URI.

Will that be a correct inference in every case? Who knows? Is it 
plausible? Sure. Is it a "protocol"? I don't know, but I'm not 
proposing it as a standard, i.e., I'm not *prescribing* it. But you 
didn't ask that. This seems perfectly programmable to me. Have a blast, 
and release it open source.

>   How best to do this? What is the
> proper protocol for publishing a dispute in RDF?  I used to work with
> ACH, the Automated Clearing House system for interbank transfers.  They
> had a dispute protocol, very well worked out, with several layers of
> automated resolution procedures before kicking the problem out for 
> human
> intervention.  Where in the current standards is our dispute protocol
> spelled out?

Nowhere. We don't have one. It would be nice to have one. It would be 
nice not to preclude various designs by being overagressive in 
specification at this point.

> What I see is that some internally consistent RDF is put on one web 
> site.
> Some more internally consistent RDF is put out on another web site.  
> Now
> if I load them both, I get an RDF graph that is inconsistent.  How does
> this become a dispute?  rather than nothing?  or garbage?

If your software interprets it as a dispute (rather than an error) and, 
I'd say personally, if the document author confirms it.

> For that matter, how do I know what was asserted?

You don't. We don't have any way of saying what's asserted on the 
semantic web. You make a leap out of specs when you assume ANYTHING is 
asserted. (No, publishing something to a website is not necessarily 
asserting it. This is *un*specified.) We do have a notion of 
coassertion and of entailment.

>  in order to know
> what to dispute?  How do I program my sw-agent to distinguish between
> some RDF that is put on a web site as a test?  or as a prototype?  
> versus
> the RDF that really matters.

You do so in an application dependent way. It's unspecified and totally 
will not be specified in this round of RDF or OWL. There's just no hope 
for it.

> How can I tell when some RDF is published
> as a comment, and when it is a claim?

Application (or system, or protocol dependent). Annotea might give you 
some notions for what it is to assert or to comment or...a bit of rdf 
*for annotea* clients and servers, but there's no generic way.

At least yet. I, personally, doubt that there will be anything near 
term. Maybe even long term.

(This is all stuff I brought up in my LC comment on section 4: Without 
a notion of when a document is asserted, the rest is a wash. Tim's 
naive protocol implicitly relies, I think, on "putting somthing at a 
namespace on the web is asserting that this is the prefer and 
determinative meaning of the defined uri")

> Could this be the protocol0, the naive protocol that Sandro and Tim
> are saying is missing?

Perhaps. But I doubt it. protocol0 thus far says nothing about 
assertion in general on the web.

> Then even as a freedom loving formalist, who would
> rather die than have my freedom of sw-speech curtailed,

Sorry, you aren't in yet. You don't know the handshake, nor do you have 
the tattoo.

> I would also have
> to say, I don't see any harm in coming up with such a protocol0, sort 
> of
> a "Quick Start Guide to Saying Things in RDF",

Uh. There's a vast difference in a "Quick Start" guide (which implies 
that there's a "long start" and a "finish" :)) and a "specified 
protocol, or even best practice, that might screw up getting things 
better later, and is likely to".

Good god, think about w3c practice and process! How on earth are we to 
come to consensus on anything like this?

Tim (or you) are, of course, free to write a "design issue" or 
whatever. And we may happily debate it casually here or even seriously 
here and elsewhere. But we're talking about imposing heavy heavy design 
and editorial and review work on lots of people.

> that explained at the very
> least, how to put some RDF on my web site and make it known that - "I 
> mean
> this.  This is something I'm asserting.  This is something that is
> disputable."

There is no standard way. Period. Though, if you publish stuff on your 
web site in rdf you should be ready for people to believe you meant 
that and dispute it.

it's a comment error for undergradaute philosophy students, in my 
experience, to get confused with a paper that *refutes* an argument, 
after having explicated it. "She argued for the view and then 
contradicted herself!!!" It's tricky, and people are prone to, well, 
getting it wrong. Before we give them advice or commands, i'd like to 
have some evidence that we've gotten it right, or at least, livable. 
Perhaps weblike, even.

Cheers,
Bijan Parsia.

Received on Friday, 10 October 2003 10:25:00 UTC