I said:
> > But since this is already standard practice, and recommended by the
> > official standard,
>
> Well, OK then. I didn't know it was already so specified.
>
> > I think we should proceed as-is. It works.
>
> It's just, it seems you end up in really hairy situations if you adopt
> this as a general practice (as opposed to a really narrow one, like
> maybe defining formulas or elements to output).
>
> For example, how does a processor know whether "re:something" indicates
> namespaced content, or whether it's simply content than happens to have
> a ":" character?
I propose that they do this trivially - if it has a potential
namespace name followed by ":", it's the namespace.
You could even say "... and there is such a namespace" if that worries you.
This is a _should_ in the specification, not a _may_.
> We'll be faced with this choice (whether to use GNames or full uris)
> probably in the metadata work, and my understanding has always been
> that it's bad practice to use QNames in this context.
>
> So we may end up with a situation where they're allowed in some places
> (say formulas) and not elsewhere. Is that fine?
Fine by me. Though I expect we'll see this kind of issue again.
This is a well-known solution to a particular problem, and
(1) nobody seems to have a better solution and (2) it's widespread practice.
I see this as an example of standards evolving to solve real-world problems,
rather than being a problem in itself.
--- David A. Wheeler
