RE: [office-comment] PRD02 Pt 1 Conformance language non-sense

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

There is at most one namespace prefix at the front of the attribute value
itself.  There is a specific set of those where the default binding is to
the OpenFormula namespace.  In all cases, the text for those specific
attributes tends to fill in the blanks.  There are other places where an
attribute value may begin with a namespace prefix and ":" for which no
default binding is specified and there is no statement about those.

Your suggestion conflicts with the intended provisions and the provisions of
OpenFormula and is definitely not intended in ODF 1.2 csprd02.

In the case of those specifically-named attributes in D.3), there are two
cases for how the namespace binding is brought into play"

 1. First, the entire syntax and semantics is determined by the prefix's
namespace binding (or the default namespace in the absence of a prefix).
This is the general case apart from the exception (2).

 2. Secondly, for a few attributes there is a specific syntax and semantics
established by ODF that allows insertion of sub-expressions whose syntax and
semantics are determined in accordance with that attribute value's single
prefix's namespace binding (or the default binding in the absence of a
prefix).  ODF does not specify any prefixing in the interior of the
attribute value, just at the beginning of an attribute value.

Whether there are further arrangements for namespaces or anything else
within the part governed by the bound namespace is completely dependent on
the definition of the particular syntax and semantics established for that
namespace and ODF is silent on the matter.  (When the OpenFormula namespace
is bound, the provisions of OpenFormula apply, in conjunction with the
applicable host-dependent behaviors.)

In the second case, there are some limitations on what the syntax of a
sub-expression might be, since it is presumably necessary to be able to
unambiguously recognize such an occurrence and determine its beginning and
end inside the overall ODF-specified attribute-value syntax.  Cf.
style:condition and table:condition in Part 1 for how we have endeavored to
improve this over earlier provisions.

I think, with regard to the clauses in section 2, it is important to also
examine the attributes that are named for a complete definition of what is
involved in these specific cases.

Apart from any concern for the suitability of these features, have I helped
clarify exactly what the provisions are?

-----Original Message-----
From: Alex Brown [mailto:alexb@griffinbrown.co.uk] 
Sent: Sunday, December 26, 2010 03:32
To: dennis.hamilton@acm.org
Cc: ODF Comments List
Subject: RE: [office-comment] PRD02 Pt 1 Conformance language non-sense

Dennis hi

> I can see that saying "the namespace prefix *in* a ... text:formula
attribute
> *value* ... " might be better, but if you look at the definitions of those
> particular attributes you will see the relevant details.

Yes, that takes it towards being intelligible/implementable, but probably
isn't enough to get us there.
 
> Does this help, and is that enough in the wording that would make this
more
> clear?

A few questions:

1. The text mentions "*the* namespace prefix", but couldn't there be many
(if a function has arguments which call functions e.g.)? Is what is really
meant here something *like* "the namespace prefix of any function contained
in the content of..."?

2. The current conformance clause mentions just syntax. But that rather
defeats the point, since surely what is desired is that these OpenFormula
formulas actually conform to OpenFormula in all regards?

[ ... ]
 
>  - Dennis
> 
> BACKGROUND
> 
> We're talking about an attribute *value* not its name.  We have a
> mechanism, used in various places in ODF, where an attribute value has the
> form of an [optional] NCName together with a following  ":" that is then
> followed by some further attribute value.  The NCName must be bound to a
> namespace and that namespace determines the syntax and semantics of
> some or all of the further attribute value.  (In the case of table:formula
> attribute values, it is all of the rest.)
> 
> This is not new in ODF, and we endeavored to improve it in ODF 1.2.
> Previously, it was somehow assumed that the ":" was part of the prefix but
> that is not true, and we modified things to say that when the [optional]
> prefix is present it must be separate from the remainder of the attribute
> value by a ":".  The omission of the Unicode Code Point is an oversight.
> 
[ ... ]