Re: [office] Namespace prefixes in attribute values

["David A. Wheeler" <dwheeler@dwheeler.com>]

Dave Pawson
> On which basis I'd advise against QNAMES in attribute values

But what, exactly, would you do INSTEAD?  We need a reasonable alternative.  It would be absurd to embed whole IRIs inside the attribute value containing the formula.  We could have a "formula-language=" attribute to an IRI, for every formula; that might be 'clean' from that perspective, but that would make the format painful to use in practice.  If formula-language="..." is a namespace value, there's no real difference than the current approach.  They're very doubtful as prefixes of the entity names, or of the attribute names, because they're always formulas (though perhaps this is the best alternative).  We could forbid alternative languages (which is what failing to spec this issue would lead to), but that makes no sense; existing docs already find this capability useful.  The logic of what we're trying to do leads to, well, what's currently spec'ed.

This is _already_ current practice; OpenOffice.org _already_ does this, as do I believe all OpenDocument implementations.  In general, I believe standards should codify existing practice, unless there's a good reason to otherwise AND a reasonable alternative.  Indeed, we should be wary of fixes that have never been tried in an implementation; often the fix is worse than the disease :-).

But if you think it's a bad idea, what would you do instead?

--- David A. Wheeler