Re: [oiic-formation-discuss] Canonicalizations galore

[robert_weir@us.ibm.com]

jose lorenzo <hozelda@yahoo.com> wrote on 06/24/2008
12:31:18 PM:

> 
> Since canonicalization of XML is not something unique to ODF, it 
> will exist regardless. Two questions might be, "to what extent
will 
> ODF standardize a canonicalization algorithm, say, on a per profile
> basis; and how much support will the standard provide for those that
> want to adapt their own canonical mappings for ODF processing? There
> are probably many little details that should be standardized. These
> details would probably also vary on a per profile basis.
> 

XML and markups based on XML have allowed variability
such that two "logically equivalent" documents may have two different
lexical representations.  Canonicalization is the process of transforming
an XML instance so that it is lexically identical to any other XML instance
that it is logically equivalent to.

So with ODF, whether there are 1, 2 or many canonicalization
transforms defined, depends on how many definitions of "logically
equivalent" are needed.

I must say that this is all very interesting, and
I can see the general use for a definition of canonicalized ODF.  But
I'm losing track of why this is relevant to profiles or to the goals of
the proposed TC in general.  Is this really needed for us to achieve
our goals?  

I could see how this would be needed if we were defining
a test for an application and wanted to ensure that a given operation in
that application yielded an ODF document that was logically equivalent
to a reference ODF document.  But there is nothing in the ODF standard
that speaks to the runtime operations of an ODF applications.  My
impression was that interop tests (the so-called atomic tests) would be
loaded into an ODF application, and their runtime renderings tested.  No
need for canonicalization there.

So what am I missing?  Where does this fit in
the Big Picture?


-Rob