Re: [oic] application scenario The Hague

[robert_weir@us.ibm.com]

Andreas J Guelzow <aguelzow@math.concordia.ab.ca> wrote on 06/01/2009 
12:23:11 AM:


> > 
> > There are three levels to be concerned with:
> > 
> > 1) Namespace
> > 2) Syntax
> > 3) Semantics
> > 
> > Level 1 is trivial and not really an interesting workshop topic.
> 
> Without agreement on this item, everything else is purely academic (and
> very misleading).
> > 

The point is that ODF 1.2 will include ODF 1.2 and the conformance clause 
already mandates a single specific namespace.  So this namespace problem 
is already solved, at least on paper.  It will be solved in reality when 
we have conforming implementations of ODF 1.2.  That is why it is trivial, 
or at least an uninteresting topic from a workshop perspective.

Or do you see something here that is a topic worth discussing at the 
workshop?  I can bring it up if there is some nuance there that I am 
missing.

> > Topics 2 &  3 are the core, and I would prepare spreadsheets in 
several 
> > combinations of namespaces and syntax so all implementations at the 
> > workshop could test their semantic understanding of the formulas, 
i.e., do 
> > they yield correct results, according to the draft OpenFormula.
> 
> Well, considering that OpenFormula is primarily a compilation of how
> current application behave with everything where they disagree being
> labelled "implementation defined" I find that quite a pointless
> exercise. Having an application that doesn't yield the "correct" result
> is simply a sign that the OpenFormula subcommittee made a mistake, since
> "correct" was defined to be the common result of all implementations.
> 
> If you disagree you could perhaps provide me with an explanation of what
> CHITEST calculates (and what this could potentially be good for), other
> than that it happens to to be what all implementations seem to
> calculate.
> 

I don't disagree with you in principle.  The point of an interop workshop 
is to identify such differences and determine whether it is an error in 
the standard or an error in one or more implementations.  It would be a 
waste of time to discuss or test things that are known problems with known 
solutions (like namespaces) or areas that are known to work perfectly. The 
point is to find new bugs and to solve them.

On the specific point of application-defined differences, I'd like to get 
a list of these areas.  Are they clearly indicated in the specification 
such that I could simply grep them?  It would be interesting to create a 
maximally non-interoperable spreadsheet using just those test cases.  One 
metric of success for the workshop would be to come out with an agreement 
among implementors to significantly reduce the number of such differences, 
by agreeing to converge on specific semantics and to clarify the 
OpenFormula specification in those areas.


-Rob