FW: [office-formula] Summary 2010-09-07 - IRI vs. URI

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

I failed to do a "reply all."  Basically, Andreas is correct about URIs
being subsets of IRIs.  

I am fixated on the image of the mapping, where there are only URIs and the
non-URI IRIs map to a subset of those.  It is the invariant URIs that are
the ones that are presumably used for namespace URIs, for
uniquely-differentiated URIs in RDF, and that (I say) should be resolved, as
URI references, to package files and fragments therein.

I suppose it is incorrect to refer to these as IRI-encoded URIs.  Maybe it
is better to call them URI-encoded IRIs, but that is still imprecise.  We
need a distinguishing term that is precisely what we need.

 - Dennis

-----Original Message-----
From: Dennis E. Hamilton [mailto:dennis.hamilton@acm.org] 
Sent: Wednesday, September 08, 2010 08:10
To: 'Andreas J. Guelzow'
Subject: RE: [office-formula] Summary 2010-09-07 - IRI vs. URI

Yes, I understand.  You're correct.

The problem is that the IRIs that can be used as URIs are the ones mapped to
URIs (the idempotent image). The IRIs that are not URIs map to a subset of
the URIs.  I am attempting to focus on the mapped-to URIs since that is what
we have to deal with in those places where URIs are called for.  So we need
to talk about IRI-encoded (that is, mapped-to) URIs as the ones we support
(though they might be presented as, entered as, and even extracted as the
non-URI IRIs that have that encoding).

While it is factual to say that, in the world of "pure" IRIs, the URIs occur
as a proper subset (and are invariant under the mapping) that also includes
URIs that we need to rule out from an implementation because of the
inconsistencies that result: %-encodings for which it is not possible to
know what [Unicode] character(s) the encodings are intended to represent and
which are not required for the mapping of non-URI Unicode IRIs to URIs.

For interoperability, it seems to me, it is necessary to restrict the URIs
that correspond to references to manifest:full-path values (and Zip
directory file names) in a way where there is no ambiguity on how to
transform an IRI to a URI first.  

The description of the mapping that has all URIs be a subset is not helpful
in that regard.  We could honor it but the interoperability that I believe
is implicit in the requirement that IRIs be supported would not be achieved.
We can, of course, give the folks what they asked for rather than give them
what it is I believe they want.  I suppose we should ask for the use case
that is expected to be satisfied by way of clarification.

 - Dennis

-----Original Message-----
From: Andreas J. Guelzow [mailto:andreas.guelzow@concordia.ab.ca] 
Sent: Tuesday, September 07, 2010 23:33
To: dennis.hamilton@acm.org
Cc: dwheeler@dwheeler.com; office-formula@lists.oasis-open.org
Subject: RE: [office-formula] Summary 2010-09-07 - IRI vs URI

Hi Dennis,

when I hear the word "subset" I understand it as a mathematical term:

IRIs form a subset of URIs means that every IRI is a URI.
URIs form a subset of IRIs means that every URI is an IRI.

This is completely separate from any consideration of mappings between
those two sets.

[ ... ]