Re: additionalType property, vs extending Microdata syntax for multiple types

Dear  Dan and all,

Just a summary and some thoughts:

1) The @itemtype attribute, if specified, must have a value that is *an 
unordered set of unique space-separated tokens that are case-sensitive, 
each of which is a valid URL that is an absolute URL*, and all of which 
are defined to use the same vocabulary. The attribute's value must have 
at least one token.

2) @itemprop must have a value that is *an unordered set of unique 
space-separated tokens that are case-sensitive*, representing the names 
of the name-value pairs that it adds. The attribute value must have at 
least one token.


3) *Property names in Schema.org are unique i.e. they have the same 
semantics* no matter which Schema.org class define/use them.

4) Microdata processor must be aware of various vocabularies  but not 
mixing them (is mixing a complexity issue?). In addition, I would like 
to emphasize that extracting Schema.org annotations from HTML is 
challenging because of various cases the HTML source can be delivered. I 
believe is a mix of DOM processing  with knowledge discovery/extraction. 
When the processor gets

<div itemscope itemtype="http://schema.org/Place">
  <span itemprop="address">3102 Highway 98, Mexico Beach, FL</span>
</div>

then it may extract a Text as address (?!) BUT it may also have chance 
to extract a nice http://schema.org/PostalAddress instance. The result 
is substantially different in terms of consumption of the semantic data.

5) To consume semantic annotations  coming from a vocabulary A distinct 
from Schema.org   one can define a mapping from A to Schema. When such 
mapping is in place then semantic data using vocabulary A can be 
consumed by a schema.org processor too.
Of course, the other way around (Schema -> A mapping) will allow 
specific semantic data processor to consume Schema.org data.
I suggest that a "standard vocabulary" ( Schema.org?) may act as an 
interchange vocabulary.

6) It is yet valid and at no big cost(?) that http://schema.org/Thing 
may define a property @sameAs:<unordered URI list> meaning the mapping 
from Schema to  different vocabularies. As Schema.org does not define 
required properties an annotation processor may or may not consider 
processing this property

Kind regards and sorry for a long email,
Adrian

On 6/19/2012 11:48 PM, Alexander Botero-Lowry wrote:
> On Tue, Jun 19, 2012 at 10:24 AM, Dan Brickley <danbri@danbri.org> wrote:
>> Thanks everyone. Lots of mail!
>>
>> I have tried to make a brief summary of some of the points in the Web
>> Schemas wiki, just a sketch of individual positions really rather than
>> a summary of the whole debate. I also started there to write up
>> details of a concrete proposal for 'additionalType'.
>>
>> http://www.w3.org/wiki/WebSchemas/additionalTypeProposal ...
>>
>> Since nobody has volunteered to lead an effort to get Microdata syntax
>> changed to support multiple types from different vocabularies, and on
>> balance after reading thru all the debate, I think we should go for
>> the new property approach.
> Funny, as I was checking the spec to comment on the vocabulary issue,
> I noticed it had changed since we implemented it:
>
> ``The item types of an item are the tokens obtained by splitting the
> element's itemtype attribute's value on spaces. If the itemtype
> attribute is missing or parsing it in this way finds no tokens, the
> item is said to have no item types.''
>
> @itemtype now does allow multiple values. Though the list is
> explicitly unordered so all of the types in the list need to be a part
> of the same vocabulary for this to be usable.
>
>
> alex
>