[Patrick Stickler, Nokia/Finland, (+358 40) 801 9690, [log in to unmask]]


----- Original Message -----
From: "ext Roland Schwaenzl" <[log in to unmask]>
To: <[log in to unmask]>
Sent: 11 December, 2002 20:38
Subject: Re: RDF typed literals and DC encoding schemes


> >
>
> > I'm not sure we're talking about the same things here. My
> > recommendations about DC defining URIs to denote terms in
> > various important vocabularies has nothing to do with RDF
> > specifically. Having globally consistent names for such
> > resources is beneficial in a far broader scope than just RDF.
>
>
> Dear Patrick,
>
> Please read the "Announcement" section of http://www.aip.org/pacs/ -
>
> I'm not sure, whether your interpretation of the re-use of classification
> codes to create DCMI owned URI's is in accordance with that section.

I don't see how it could possibly conflict with it.

Again, if DCMI minted URIs to denote PACS codes, how does that
infringe on PACS ownership of those resources. All DCMI is doing
is providing another name in the form of a URI for each PACS code,
and defering all interpretation of that code to the PACS authority.

If PACS were to define its own URIs for its codes, then DC users
could just use those, but if they don't, how can it possibly be
wrong to define your own URIs.

This is a fundamental right of the Web and Semantic Web, to mint
whatever names you want to denote whatever you want. Sure, you
could be held liable for what you say about those things, if
your statements conflict with those of the owner of the thing,
but simply creating your own name to denote the thing is a right
(and habit) that folks have had since the dawn of language.

If I decide to call my dog a buuajsswiwkaflat, I can do so. Other
folks may choose not to use that name, and lots of folks will
probably not know what I'm talking about (without some context),
but no'one can tell me that I *can't* mint my own new name for
that species of animal.

I think you are combining the act of minting a name (URI) with
the act of defining/describing the resource denoted by that name.
These are two very distinct things.

> Can you give convincing legal evidence, that under whatever
> jurisdiction there is no risk for DCMI to re-issue classification codes
> as URI's under it's own authority?

See the "buuajsswiwkaflat" example above. Or better yet, I challenge
anyone to prove that the following URI, that I now mint myself,
based on a domain I personally own, to denote you personally, is
in any way illegal by any reasonable interpretation of any set of
laws:

   <voc://infospring.com/people/12345> dc:description "Roland Schwaenzl" .

I may use that URI to talk about you as much as I like.

Now it may be the case that something I say about you might constitute
liable or slander or such, and you may have every right to sue me for
something that I might say about you, but simply minting the above URI
to denote you in my statements about you is my right and in no way
violates your own rights.

So, in the same way, if some organization does not themselves define
URIs to denote resources they own and which are useful to DC users,
DCMI should feel no guilt in defining URIs to denote those resources.

As for "reissuing" codes, if a code defined by some other agency is
used as a component of a URI, such that the meaning of the code is
not changed, how is that an infringement on the owner of that code.

Does the URN urn:issn:123... infringe upon the owner of that ISSN
code, by being a component of a URI that they didn't define? I think
not.

> -------
>
>
> xsi:schemaLocation -
>
> One could have an xml-element preceding rdf:RDF,
> which carries this information -
>
> I'm not sure, whether RDF/XML parser generally be happy
> with that - as they might conclude on RDF/XML not enclosed
> by rdf:RDF .
>
> xsi:type cumbersome as well. It's use typically will result in
> RDF/XML syntax errors.
>
> Seems to me one of the reasons, which make it difficult to
> create XML schema for XML dialects as RDF/XML, which make
> their content models in dependency of attribute configurations.
>
> One could think this is a problem of XMLSchema rather than RDF/XML -
> but ...

I think that the entire xsi: vocabulary is a mistake and reflects
a certain arrogance of XML Schema over the broader XML community.

I don't expect that any RDF/XML parser should have to correctly
interpret any xsi: vocabulary. RDF uses XML for its serialization,
not XML Schema.

I hope that in future editions of the XML Schema specs, the xsi:
vocabulary would be deprecated.

> ---------
>
> When it is true, that RDF effectively (!) thinks about datatyping just
> as providing a pair of objects (x, y) ; x a character sequence not using < > </ >
> and a URIref - why the drafts
> go for a literal2value map at all?

Because the lexical form is just a means to an end, and that end
is the datatype value.

Furthermore, the L2V mapping is essential because the mapping
is N:1 where N>=1 -- i.e. there may be more than one lexical form
that maps to the same value, and thus, simple comparison of
typed literals is insufficient to determine equality.

E.g.  "1.0"^^xsd:decimal == "1"^^xsd:integer == "000001"^^xsd:byte

> If it is completely to the application to
> make any sense of such a pair, one
> could give "RDF datatyping"
> a much simpler wording with less philosophy in it.

It is important to define the foundational machinery of
datatypes so that interpretations of typed literals are
constant accross applications. E.g. if we didn't state
explicitly that the L2V mapping must be N:1 then it would
be possible to define a datatype that was ambiguous, where
the same lexical form could map to different values, and
that would hardly be a good thing.

Trust me, we worked very hard at making the RDF datatyping
solution as simple, minimal, and generic as possible while
still getting the job done. If there's something there, it's
necessary.

> As the Primer is doing: In partice one can simply use any URI to
> indicate a datatype.

Well, a datatype is denoted by a URI, and RDF datatyping is
designed to work for any datatype that conforms to the
characteristics as defined for members of rdfs:Datatype,
and not necessarily XML Schema datatypes.

Patrick