RE: LANG: need to CLOSE Issue 5.6 Imports as magic syntax

Jerome, I'm replying once here to your two emails [1] and [2].

> >pps In case there's no consensus, note I would not be opposed to the
> >(at that point, reasonable) "leave import after v1" proposal
> >that DanC and Jim are advocating.
>
> Would this practically mean that users will have to put all their
> assertions in the same <rdf:RDF>  </rdf:RDF> until the next version?
>
> (this is a real question, please answer, but I carry on).
>
> I guess not. This is my understanding that you still can refer to
> objets by someURI#Foo and the interpreter will be in charge of
> computing the closure for getting the meaning (*). Right?
> That's the implicit import.
>
> Is this the alternative that we are facing?
>
> (*) The only objection I find to this way of doing is that, I still
> do not really know which a URI should allows us to locate a
> definition (but I can sleep with that). The objection that Web people
> might make to both kind of import is that content is negociated, and
> thus you never now what will really be imported in another context.

Well, you touched an important issue in itself. This issue has been touched on in [3]:
now, although there (and in general) I wouldn't advocate a process model that made this way of
using URIs mandatory (ie, forcing people to have URIs that denote their physical location too), cf [4],
it's certainly possible and welcome (like I think you are suggesting here) to act under that
process model in the rfc2119-MAY flavor.
The MAY flavor is also the one one could nicely adopt with the rdfs:seeAlso, this way we could
smoothly close the import issue. rdfs:seeAlso or specific other explicit import relation has one advantage over the URI-as-location
model described above, though:
a) the URI-as-location might cause extra burden on the system (when to check? likely, always... and if so, many requests might have
been done that are not necessary), and
b) in any case, sometimes you need to go beyond the URI-as-location (specifying a different location): so, you'd need more in any
case (like the rdfs:seeAlso)

So, well, we'd be careful on whether to also introduce the URI-as-location model, and better analyze its possible computational and
architectural overhead.

> Now to consider the point made by Massimo:
>
> At 15:07 +0100 07/11/2002, Massimo Marchiori wrote:
> >1) [major] when transporting/manipulating RDF, you risk to lose the
> >import information.
>
> This can occur if you consider that once you've imported the
> definition of someURI#Foo, you are entitled to call it "Foo" in your
> RDF/OWL. Is this the issue?
>
> Would someone be opposed to have an "import" construct + the
> obligation to keep the someURI# ? This should solve this problem at
> least. Import would just mark the endorsement of the imported content.

Well, no, I was just talking about transporting as RDF triples (I thought the latest incarnation of the proposal implied the import
is not represented in the graph but just at OWL XML serialization level, which fortunately is not the case). But the point you
mention is in fact related to the second one I cited in that email (linking the entailment with the Web structure).
Rather than talking abstract, it's better to see the subtle consequences of that point (that I haven't explicitated using email so
far, just hinting at them in some teleconf some time ago), so it's better I first read Pat's latest wording on the import ([5]), and
see whether they apply there (I think so, but can't be sure until I read the wording (sorry but it's Sunday nite here now ;). Of
course if they don't apply (unlikely but who knows), well, I might be the first to defend that wording to the teeth :)

-M


[1] http://lists.w3.org/Archives/Public/www-webont-wg/2002Nov/0085.html
[2] http://lists.w3.org/Archives/Public/www-webont-wg/2002Nov/0084.html
[3] http://lists.w3.org/Archives/Public/www-webont-wg/2002Sep/0456.html
[4] http://lists.w3.org/Archives/Public/www-webont-wg/2002Oct/0312.html
[5] http://lists.w3.org/Archives/Public/www-webont-wg/2002Nov/0099.html

> -----Original Message-----
> From: Jerome Euzenat [mailto:Jerome.Euzenat@inrialpes.fr]
> Sent: Thursday, November 07, 2002 5:54 PM
> To: Massimo Marchiori
> Cc: www-webont-wg@w3.org
> Subject: RE: LANG: need to CLOSE Issue 5.6 Imports as magic syntax
>
>
> Hello,
>
> 	I need some clarification here.
>
> In his message (RE: LANG: need to CLOSE Issue 5.6 Imports as magic
> synt) of 07/11/2002, Massimo Marchiori wrote:
> >pps In case there's no consensus, note I would not be opposed to the
> >(at that point, reasonable) "leave import after v1" proposal
> >that DanC and Jim are advocating.
>
> Would this practically mean that users will have to put all their
> assertions in the same <rdf:RDF>  </rdf:RDF> until the next version?
>
> (this is a real question, please answer, but I carry on).
>
> I guess not. This is my understanding that you still can refer to
> objets by someURI#Foo and the interpreter will be in charge of
> computing the closure for getting the meaning (*). Right?
> That's the implicit import.
>
> Is this the alternative that we are facing?
>
> (*) The only objection I find to this way of doing is that, I still
> do not really know which a URI should allows us to locate a
> definition (but I can sleep with that). The objection that Web people
> might make to both kind of import is that content is negociated, and
> thus you never now what will really be imported in another context.
>
>
> --
>   Jérôme Euzenat                  __
>                                   /      /\
>   INRIA Rhône-Alpes,            _/  _   _   _ _    _
>                                /_) | ` / ) | \ \  /_)
>   655, avenue de l'Europe,    (___/___(_/_/  / /_(_________________
>   Montbonnot St Martin,       /        http://www.inrialpes.fr/exmo
>   38334 Saint-Ismier cedex,  /          Jerome.Euzenat@inrialpes.fr
>   France____________________/                Jerome.Euzenat@free.fr
>

Received on Sunday, 10 November 2002 19:26:47 UTC