2009/3/4 Dennis E. Hamilton <>:
> The following remarks are all speculative, because we don't know whether
> anyone will stub their toe on these or not.
A little Dennis, just a little :-)
> ANALYSIS
> ========
>
> 1. We still have a problem with foreign attribute values on existing
> attributes. When such an attribute has no default value (the defined
> fall-back) and the attribute is required, the document would be
> unacceptable. I don't think your proposal addresses this case anyhow, but I
> do notice that language being introduced for it has this hole. I also note
> that those places, such as table:formula, where all of the values are
> foreign in ODF 1.1 [and presumably still permitted in ODF 1.2 (extended)
> documents], the fallback from only one something to only nothing is pretty
> stark.
As a generality, I'm not in favour of backwards compatibility.
I'd suggest 1.1 'extensions' be treated as any other 'extension'
>
> 2. One problem with the use of attributes only is that there can only be one
> occurrence of an attribute of the same name per element. There is no
> provision for multiplicity, repetition, etc.
Same local-name, different namespace?
Are they 'different'?
>
> 3. One way to deal with situations like (2) is to use the ODF counterparts
> of <div> and <span> or (hack, cough) the ODF bookmark provisions that allow
> hierarchy to be transcended in the manner exploited by the RDF Metadata
> feature.
Scenario. You are an implementor. You come across a valid ODF element
with foo:bar="The quick brown fox" as an attribute.
Would it make sense to treat this as content? I don't think so.
I'd suggest as a 'maximum', the presentation layer is asked to inform
the reader that such foreign content exist.
More practically, as a document, report that it contains an extension
and leave it at that.
>
> 4. I said this is all speculative, because we don't know whose situation
> would be thwarted by the constriction.
The only applications that can reasonably process extensions are those
that comprehend the element/attribute/namespace combinations,
probably because that application generates them.
That makes sense, in a local, non conformant usage which I see
as doing no harm when used in that way.
That's how good features can be trialled for adoption by
the standard.
regards
--
Dave Pawson
XSLT XSL-FO FAQ.
Docbook FAQ.
http://www.dpawson.co.uk
