Re: Change on typeof? (Was Re: The processing rules for property/rel)

On Nov 1, 2011, at 19:35 , Gregg Kellogg wrote:

> On Nov 1, 2011, at 6:06 AM, Ivan Herman wrote:
> 
>> Gregg,
>> 
>> I just try to go down the line of concentrating on @typeof; maybe that reaches the same goal as what we wanted, but makes a change in RDFa that might be, possibly, cleaner. This is not fully worked out yet...
>> 
>> Let us consider changing the behaviour of @typeof such that:
>> 
>> 1. if (@rel/@rev is present, but no @resource/@href/@src) or 
>>     (@property is present, @rel/@rev is not, and no @resource/@href/@src)
>>   then 
>>      - @typeof creates a blank node
>>      - that blank node is the current_object (ie, the one going down the chain)
>>      - is the subject of the type triples
>> 2. if @resource/@href/@src is present _as well as_ @typeof, then no blank node is generated, and the subject of the type triples is the IRI resource of @resource/@href/@src
>> 
>> Both of these rules are different than what we have now; all other rules for @typeof remain unchanged. The advantage is that there is no difference between @rel/@rev on the one hand and @property on the other hand with respect to @typeof. Of course, what I called the 'minimalist' change for @property, as you described on the wiki, remains valid, too.
> 
> I think we break a lot if <div about="foo" typeof="Bar"/> does not assign the type to <foo>. However, saying <div typeof="Bar"/> or <div typeof="Bar" resource="baz"/> assigning the type to the object would be interesting.

As I said: the rules above are an addition to the current rules, and they are silent in cases there is no @rel/@rev. What this means is that in all three examples the current rules apply:

<div about="foo" typeof="Bar"/>

sets the type for 'foo', which will be the subject down the line

<div typeof="Bar">

creates a new bnode of type 'Bar', which will be the subject down the line

<div typeof="bar" resource="baz"> 

will set the type for baz and that will be the subject down the line.

> 
>> What we get is:
>> 
>> <div about="relV" rel="t:foo" typeof="t:Bar"><span property="t:yip">Yes</span></div>
>> =>
>> <relV> t:foo [ a t:Bar;
>>           t:yip "Yes" ] .
> 
> With my interpretation, presuming that @rel/@rev change to reference _new object_ and not _new subject_, and this would create the following:
> 
> <> t:foo <relV> .
> <relV> a t:Bar;  t:yip "Yes"  .

Yes, indeed. You offer an equal interpretation to @about and @typeof in this respect, whereas I only touch at @typeof. My approach was to find the minimal set of changes that would cover a large number of use cases. I am not convinced that mixing in @about into the equation makes sense.

I also must admit that @rel applying to @about in terms of object is counterintuitive for me. @about should be, essentially, on the subject (as the name of the attribute also suggests) and @resource (or @href) is the intuitive attribute for an object. But that may be only me.

> 
> If @rel retained it's current interpretation, there would be no resource for @rel to operate on, and it would do the following: 
> 
> <relV> a t:Bar;  t:yip "Yes"  .
> 
>> <div about="relP" property="t:foo" typeof="t:Bar"><span property="t:yip">Yes</span></div>
>> =>
>> <relP> t:foo [ a t:Bar;
>>           t:yip "Yes" ] .
> 
> This should be consistent with the first interpretation above:
> 
> <> t:foo <relV> .
> <relV> a t:Bar;  t:yip "Yes"  .

I guess what you say that there should be consistency between @rel and @property. I agree. So if the behaviour of @about is also changed, that this should also change, but I think what I describe _is_ consistent...

> 
>> Both by virtue of #1 above. The reason for the #2 rule above is that if I have
>> 
>> <div about="relV2" rel="t:foo" typeof="t:Bar" href="http://example.org">
>> 
>> Our current rules would assign a type to <relV2>, but this is really not consistent with the fact that, without @href, the typing goes to the 'right hand' of @rel, so to say. With that modification, what we get is:
>> 
>> <div about="relV2" rel="t:foo" typeof="t:Bar" href="http://example.org">
>> <span property="t:yip">Yes</span>
>> </div>
>> =>
>> <relV2> t:foo <http://example.org> .
>> <http://example.org> a t:Bar;
>>   t:yip "Yes" .
>> 
>> <div about="relP2" property="t:foo" href="http://example2.org" typeof="t:Bar">
>>  <span property="t:yip">Yes</span>
>> </div>
>> =>
>> <relP2> t:foo <http://example2.org> .
>> <http://example2.org> a t:Bar;
>>   t:yip "Yes" .
>> 
>> Thoughts?
> 
> I would favor @typeof binding to either @about or @resource/@href/..., but not both. On it's own, having @typeof describe the type of the newly created object makes most sense to me, if we decide that some change is necessary.


Well... what you propose would destroy the <div about="..." typeof="..."/> idiom although, in the absence of @rel or @property this would be intuitive for me...

Ivan



> 
> Similarly, binding @rel and @property to the object, rather than the subject, produces what I believe to be the most sensible results, but does create a compatibility issue for @rel, which is why I only suggested it for @property.
> 
> My proposal had @property bind to _parent subject_ instead of _new subject_, but if we make these changes, I would support binding it to _new object_, so that the recommended pattern would be:
> 
> <div property="t:foo" typeof="t:Bar href="http://example2.org"><span property="t:yip">Yes<?span></div>
> 
> which would create
> 
> <> t:foo [ a t:Bar; t:yip "Yes" ] .
> 
> @href without an @typeof would not create a new object.
> 
> Gregg
> 
>> Ivan
>> 
>> 
>> 
>> ----
>> Ivan Herman, W3C Semantic Web Activity Lead
>> Home: http://www.w3.org/People/Ivan/
>> mobile: +31-641044153
>> PGP Key: http://www.ivan-herman.net/pgpkey.html
>> FOAF: http://www.ivan-herman.net/foaf.rdf
>> 
>> 
>> 
>> 
>> 
> 
> 


----
Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf