W3C home > Mailing lists > Public > public-vocabs@w3.org > February 2013

Re: Re-usable bits/facets was: Re: aggregateRating and review under /Person

From: Dan Brickley <danbri@danbri.org>
Date: Sat, 9 Feb 2013 18:00:09 +0000
Message-ID: <CAFfrAFov7hu5JBbmsuW9sHovzrnPQnycEJR63Ls9-vUbeE4XCw@mail.gmail.com>
To: Martin Hepp <martin.hepp@ebusiness-unibw.org>
Cc: Karen Coyle <kcoyle@kcoyle.net>, "public-vocabs@w3.org" <public-vocabs@w3.org>
On 9 February 2013 11:35, Martin Hepp <martin.hepp@ebusiness-unibw.org> wrote:
> Hi,
> just from the top of my head: The data model behind schema.org supports two ways of using a property with multiple types:
> First, properties are inherited along the type hierarchy, so subtypes share all properties with their supertypes.
> Second, one can define the domain of a property to be the union of multiple types. So for instance, if the property xyz should be applicable with two types T1 and T2, it is not necessary to define a type T3 which is a common supertype of both. Instead, one can directly specify a list of types for which this property shall be available.
> That is a nice deviation from the OWL semantics, where you have to define an implicit class for sharing a property.
> This feature is currently visible only when you directly work on the schema and there is to my knowledge no tool support for managing this.

Also worth mentioning (and again not captured by many existing
formalisms and tools), ... schema.org is a gradually evolving system.

For example in http://schema.org/Movie we have 'actor' which takes
'Person' values. No other class is *currently* listed. However we do
gradually make small changes here and there. So to describe the TV and
Movie contributions made by the dog-actor known as Lassie, you could
imagine we might someday allow a class like 'Animal' as a second range
of actor. One way of knowing when it's worth making such changes is to
be permissive rather than strict about data that goes beyond the
current formal schema... either by using hypothetical / extension
types, 3rd party vocabularies, or stretching range and domain rules
when it seems sensible. It's a balancing act; too much of this and the
data becomes chaotic; too little, and we're rigid, frozen in our
current state, making incremental improvements difficult.


ps. I consider Lassie a Person so schema is fine imho...
Received on Saturday, 9 February 2013 18:00:37 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:48:53 UTC