W3C home > Mailing lists > Public > public-vocabs@w3.org > November 2012

Re: proposal for new classes Activity and Action

From: Dietrich Schulten <ds@escalon.de>
Date: Sun, 04 Nov 2012 17:50:33 +0100
Message-ID: <icpek1xtn4dxlmp01636j28k.1352039597432@email.android.com>
To: public-vocabs@w3.org
Hi Daniel, 

Can you clarify a few things for me,  please? I am quite new to ontologies and vocabularies,  so maybe I missed a few points here. Or maybe my questions turn out to be valid. 

I am currently trying to understand the relationship between your proposal and activitystrea.ms. 
First of all,  activitystreams have a concept of a target in addition to the object of the verb. This accounts for the fact that many verbs allow or even need more than one object in a sentence. E.g. "I give this present to my son" ,  or,  "I call this animal a mammal".

You say "The Target does not correspond to any one property, and may be represented either as a property of Activity/item or a property of Activity/result depending on the verb."  This is not clear to me. How would you express "Geraldine posted a photo to her album"? Can the activity post have more than one item? If not,  what is the item and what is the result in the Geraldine sentence? 

Second question, how are activitystreams verbs related to your proposal? You use verbs for several things. 

- Activity.action. Can be a Url or an Action. 1 Do you see any relationship with activitystreams verbs here? 2 Is an Activity.action url a shortcut for Activity.Action.url? 
- Action.name: second person imperative verb (same as activitystrea.ms?) 
   Action.url: executes the action. A simple url will usually not be sufficient to execute an action. One would need to know if you do a post, a put or delete on the url in order to execute. A simple get should not change anything and can hardly execute any actions. An action might also require several url requests. Maybe the url should rather represent the outcome of the action,  some kind of cancel receipt or a ticket or an uploaded image. After all, in an activity stream one would mostly be interested in what was accomplished by the activity rather than re-doing it. Or the url simply gives you the schema.org description of the action. 
   Action.undoAction: in many if not most cases, a simple undo is not possible for an action, rather you need a separate activity (delete a posted item,  cancel a ticket purchase) to revert to the previous state of the world. A simple url will sometimes not be sufficient to do such a thing, one needs at least an http verb (delete) or an entry point to a more complex activity,  such as a cancel order form. 
- Action subtypes: seem to be meant to recognize an Action not only by its name but also by its type. However,  the name and subtype can be inconsistent, there might be an unwanted redundancy here. And again,  how are the Action subtypes related to activitystreams verbs?

- Activity.result sounds very interesting to me. I came to your proposal from a very specific background: how can a REST Api describe possible activities in such a way that a machine client can find out how to accomplish that activity if it only knows it wants to perform an activity in a certain context. Short answer: it must follow its nose to the desired activity result by inspecting linkeddata information attached to rels in the context. I need a way to express: following this link will eventually get you a representation of a desired activity result. I need to experiment with your proposal a bit to see if it fits that purpose. 

Best regards, 
Dietrich 
Received on Sunday, 4 November 2012 23:03:06 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Sunday, 4 November 2012 23:03:06 GMT