[uf-new] collection-design-pattern proposal

[Manu Sporny]

Danny Ayers wrote:
>> On Apr 22, 2007, at 10:12 PM, Manu Sporny wrote:
>>
>> > I propose that we use this collection-design-pattern mentioned
>> > above for
>> > specifying all collections due to the following advances over all
>> > of the
>> > other proposed methods:
> 
> Nice idea, but I believe it's redundant. HTML already has collections,
> ordered and unordered.

Yes, you are correct according to your definition of a collection. The
problem I'm describing is a little deeper than that...

Apologies, I wasn't clear enough when it came to defining what a
collection was... here are the definitions:

collection: a non-localized, unstructured grouping of items.
list      : a localized, structured grouping of items.

A list is a subtype of a collection - the ontology would look something
like this:

+ - grouping
  + - collection
    + - list

HTML does provide lists - but that is only a sub-set of the 'grouping'
problem.

> (which re-raises the question of whether class="xoxo" is desirable)

xoxo is desirable when you have two lists of items listed in a single
Microformat <div>. If you want the first one included in the grouping,
but don't want the second one - then you need something like xoxo.

In other words - we want to be explicit, not implicit, about what is
included in a uF and what is not.

> Other relations can be expressed:
> 
> <ul>
>   <li id="xthing"><a href="#ything" rel="neighbour">an item</a></li>
>   <li id="ything">another item</li>
> </ul>
> 
> (probably not perfect, but I think it could be made to work)
> 
> Specific definitions of the interpretation can be provided using a
> metadata profile, e.g. by providing an RDF mapping via GRDDL.
> 
> What more is needed?

The following is needed beyond the examples you gave:

 - The ability to relate items that are not local to one another in an
   HTML page.
 - The ability to relate items on an unstructured basis.

-- manu