Re: Question about tags withen tags

[Aaron Swartz <aswartz@swartzfam.com>]

Mark Nottingham <mnot@mnot.net> wrote:

> An alternative, to-me intuitive solution is to use namespaces, as in
> <rss:title><html:i>italic</html:i> in the title</rss:title>
> so that the rss namespace is dispatched to the intermediary processor
> (in this case, the rss client), while the html namespace is given to
> the end client (browser).
> 
> Unfortunately, this still requires out-of-band knowledge about the
> namespaces and who should process them, and the granularity of that
> dispatch.
> 
> I'm just getting my head around this...

Yup. It's an interesting issue. So while technically the title elements are
defined as plain text, making all of this stuff speculative, let's indulge
ourselves for a moment here:

If we think of this stuff as layers, with different portions of the
information meant to be processed by different people, we can see it as sort
of unwrapping a package, where each system decodes its part before passing
it on to the next. The problem with this is that it's "dangerous" -- systems
that aren't safe enough to check for strange things (like nasty JavaScript,
or the more mundane blink tag) will end up doing things they don't want.

The alternative is, as you say, namespaces, but the systems need to know how
to deal with them. And, if we're adding new namespaces all the time, the
systems need to be "smart" enough to figure out how to use them
automatically. So, we could put some sort of description file at the
namespace that explained how to deal with it (we could probably break it
down into several classifications) and then have the processor fetch that to
decide how to process it.

Of course, this is all just insane noodling, and it shouldn't be used with
the current RSS, but it's interesting to think about.
-- 
[ Aaron Swartz | me@aaronsw.com | http://www.aaronsw.com ]