Re: Atom extensibility

[Danny Ayers <danny.ayers@xxxxxxxxx>]

On Fri, 07 Jan 2005 07:48:25 -0800, Tim Bray <Tim.Bray@xxxxxxx> wrote:
> 
> On Jan 7, 2005, at 6:54 AM, Henry Story wrote:
> 
> > The question is exactly how should the interesting discovery that an
> > Atom document is an RDF document [Ø] be used to fulfill the charter
> > requirements on extensibility?
> > Clearly it could completely underpin the extensibility framework, and
> > so allow the Atom charter to be successfully fulfilled.
> 
> I think that the charter requirements on extensibility will be filled
> adequately with PaceExtendingAtom.  I think they would be filled still
> better by adopting PaceMustUnderstandElement, but apparently others are
> unconvinced.  

The charter requirements could be fulfilled in any number of ways, not
all of them useful.

I think PaceMustUnderstandElement may well be a good idea.
PaceExtendingAtom does shield Atom core data from unwanted
interference from foreign elements, but that shielding can still
happen alongside partial understanding.

Extensibility via a mapping to RDF seems to me to add a
> lot of complexity (most people have never bothered to learn graph
> theory) without any real benefit. -Tim

There are feeds in RDF from the blogs of millions of users that
haven't learnt graph theory.
Given Henry's proposal, the person that uses RSS 2.0 today would
notice no additional complexity. The only people that would need to
know a little about RDF would be those that wish to develop
extensions. All they would really need to know are a few syntax rules.

On the other hand, if extension markup can appear anywhere (as in
PaceExtendingAtom) application developers will have to deal with the
complexity of handling every extension completely on a completely
case-by-case basis.

Extensibility via a mapping to RDF will happen, whether it appears in
the specifications or not. Whether it can be done consistently across
extensions will depend on how consistently those extensions are
expressed in Atom.

I personally would prefer an RDF-oriented approach in the core, as I
think RSS 1.0 demonstrates its utility. But if there is obtuse
antagonism towards RDF then a reasonable compromise would seem to be
some having some simple structural syntax constraints that extensions
should follow.

Cheers,
Danny.

-- 

http://dannyayers.com