RE: [office] 2009-01-08 Single ODF 1.2 metadata proposal - URLS OF PACKAGES

["Dennis E. Hamilton" <dennis.hamilton@acm.org>]

I need to clarify something around giving an ODF document's package an
absolute URL.

1. Using that URL as the leading part of a URI that ends with path segments
and a resource inside of the package is not understandable as a reference to
the resource within the ODF package, since we have not defined a URI scheme
that does that (and we resist using those of others that do).  Indeed, we
have forbidden URLs being used this way from within an ODF package.  

2. Using that URL from within the package is brittle because when a copy of
the document is somewhere else, that URL no longer makes sense.
Furthermore, it is silly to require the document to be updated to reflect
the new location (and in any case, might not be permissible for signed
documents, whether using the ODF 1.2 DSIG or another scheme that requires
the package to be kept intact).

3. It is for that reason that I suggest that some other device be used to
identify the document instance (at the time of production) so that the
relationship between that particular document and the metadata in and about
it can be bound without requiring the document to be at some unambiguous,
absolute location.  

4. I can imagine a number of corner cases and conflicts with URI rules (with
or without package section 2.6), so this would have to be worked out and
specified very carefully.  I saw an example in the RDFa specification that
might be helpful as a hint.  I also agree that extracting the metadata will
require something to be done about those URLs because they have no
translation as into-package URLs for use from outside of a package until
someone comes up with a URI scheme for such purposes (again, not mentioning
the scheme of which we may not speak).

 - Dennis

PS: Whether or not Sir Tim wants us to have resolveable URLs, we don't have
a URI scheme that works for the interiors of ODF packages using existing
absolute URL schemes.  Unless someone creates one (including us) and we rely
on or include it in the ODF specification, I don't think we are doing
anything to overcome that particular limitation.  Maybe your pkg:hasPart
property is a limited way to accomplish that, but it is not at all clear
from the current specification how that indirection gets us past the
previously-unsolved problem.

PPS: I don't think I need a call from some RDF experts to explain this to
me.  There should be enough information in the specification where I could
find the authoritative information in a fairly direct way.  I do believe
that I am sufficiently "skilled in the art" to be able to apply such
authoritative material, when referenced in precise ways.  

PPPS: I am not claiming that URLs are bad and that resolvable URIs are not
valuable for RDF and the Semantic Web (though I don't think that is ODFs
problem to solve).  I claim that anchoring a persistent document that is
interpretable independent of the location of its persistent form (unlike a
web resource that might not even have a persistent existence) is a special
and important problem and it takes a different kind of URI to deal with
that.  In addition to that, there is the problem of wanting to refer to an
element of an item within a Zip package and having internal relative
references and some sort of external reference that works (and can be
derived from the internal reference, hopefully).



-----Original Message-----
From: Svante.Schubert@Sun.COM [mailto:Svante.Schubert@Sun.COM] 
http://lists.oasis-open.org/archives/office/200901/msg00085.html
Sent: Sunday, January 11, 2009 13:05
To: dennis.hamilton@acm.org
Cc: 'OASIS Office'
Subject: Re: [office] 2009-01-08 Single ODF 1.2 metadata proposal

[ ... ]

Dennis E. Hamilton schrieb:
http://lists.oasis-open.org/archives/office/200901/msg00084.html
> [ ... ] 
> 2. The idea that the ODF Document must have a physically-resolvable URI
goes
> too far, especially when the example is of a location on the hard drive of
a
> computer for which there is no web-resolvable URI in evidence. 
>   
I believe you think that if you make an URL based on your hard disc 
would not work.
Someone always has to make URLs in the context of his 'network' / 
'world' / 'domain'.
This might be a computer, a LAN, the Internet or something else.
If I identify all documents on my computer, it is fine as long I make 
statements in this domain and do not leave it (mix it with other 
computer domain).
In a similar manner Internet index are being created and they work fine.
>    2.1 The difference between (copies of) document instances and some
> understood URI that can be used to refer to features of any such
(unaltered
> copy of an) instance needs to be allowed for (and this is often recognized
> in RDF and demonstrated in W3C RDFa examples, especially but not limited
to
> the reliance on Dublin Core concepts). Examples of such features are an
ISBN
> number, the name of the author, bibliographic information about it,
> bibliographic information about something mentioned or discussed in it,
> assertions about what someone says about it, assertions about what is
being
> said in the document about what someone says about something, such as in a
> book review or literary critique, etc.  
>   
Perhaps you are confusing the need of an URL (which comes along with the 
persistence of the format (office document), remember ODF is the 
persistent file format of Office applications) with the first appearance 
of a lack of an URN.
Please note that it is still possible to provide an URN in addition with 
the URL (using owl:equal)
URL are the base of a web and that's what the Semantic Web should be about.
>    2.2 The use of URIs as concept identifiers that are never meant to be
> resolved is critical in the working of RDF and is the underpinning of the
> identification of relations and predicates and properties in RDF.  In many
> respects, the document instance is a different entity than a particular
> location where a copy of the instance might be located in a physical
> repository or on the Internet.  It seems clear to me that these prevalent
> uses of RDF (illustrate in examples of RDFa) should not be constrained for
> RDF annotation of ODF documents.
>   
Yes, that is what we (me included) had previously thought so.
The creators of RDF, the Semantic Web Interest (SWI) Group, Dan Brickley 
(W3C SWI chair) and especially Tim Berners Lee, who took an hour to 
convince me after my presentation, that I was wrong.
Convinced me of something different.

Again in a certain part you are right, but you are completely wrong in 
ignoring the advantage of an URL. If someone is annotating (adding 
metadata) to a network resource, the identifier should be the location 
to make the web possible.

[ ... ]