- From: Bill Looby <bill_looby@ie.ibm.com>
- Date: Wed, 18 Feb 2015 10:26:31 +0000
- To: Social Web Working Group <public-socialweb@w3.org>
- Message-ID: <OF23CD862B.94F3D45E-ON80257DF0.0037D7FD-80257DF0.00395DEC@ie.ibm.com>
On the vocabulary - If we define a vocabulary we imply that processors must honour this, including the hierarchy - This may become especially relevant when looking at federation and the flow of related information (e.g. a PubSubHubbub approach) - An alternative mentioned was to go completely flat (though this seems unnatural in terms of json-ld support) Rgds, -Bill. From: "Social Web Working Group Issue Tracker" <sysbot+tracker@w3.org> To: public-socialweb@w3.org Date: 17/02/2015 19:17 Subject: social-ISSUE-12 (ActionStructure): Action Types Structure and Processing Model [Activity Streams 2.0] social-ISSUE-12 (ActionStructure): Action Types Structure and Processing Model [Activity Streams 2.0] http://www.w3.org/Social/track/issues/12 Raised by: Erik Wilde On product: Activity Streams 2.0 in today's telco we discussed the problem of how to reliably handle actions and the fact that there possibly is a structure to them. for example, if (as is currently the case) the spec says "a like is a respond", what does that mean for producers and consumers? - are producers free to assume that if they create a "like", it will be interpreted as "respond" as well? if that's not the case, are producers encouraged to explicitly label a "like" *also* as a "respond". - are consumers free to assume that if they ask for AS "respond", they will also get "like" and other subclasses of activities? the current spec language is vague, and james has said that using the activity structure is optional. in this case, neither producers nor consumers can count on the structure being implemented, and as a result, if they want to use the class structure reliably, they have to do the "double labeling". this is what we currently do in our AS1 implementation: you would always label a "like" as a "like" and then use an additional property (that we made up) to *also* say "and btw, this like is a digg". if you label your "digg" as a "digg" on the verb level, it is not, by definition, a like. that system may not be great, but it results in predictable behavior and it is easy to implement. AS2 needs at the very least to be well-defined, which it currently is not. also, this question should be answered for both cases: - activities from the core vocabulary and how to process them (the "respond" and "like" case). - activities from extension vocabularies (what we discussed on the email list as a proprietary "floop" which claims to be a "like").
Received on Wednesday, 18 February 2015 10:27:12 UTC