Re: CCXML Implementation Report: asserts 714 and 715 - ISSUE-670

Paolo,

The proposition sounds reasonable for me. I believe that now this 
property is unambiguous and more useful.

I just want to note that in cases (1) and (2) it is responsibility of 
the application (script) author to include the sessionid property in the 
<send> element. I don't think that a platform should be responsible for 
checking that the property is included since the property is optional. 
Please consider this when changing the test.

Regards,
Petr


On 2.7.2010 10:49, Baggia Paolo wrote:
> Petr,
>
> We are re-discussed the issues around 'sessionid' on 'ccxml.exit', see below the proposition to change in the specification the definition of that property to make it useful. In attachment, a picture to help the clarification.
>
> Please check if it sounds reasonable. This means that both spec and the test will be change accordingly.
>
> Regards,
> Author of CCXML-IR Plan
>
> ------
> Starting from ISSUE-670 we think that a clarification within the CCXML spec is required on sessionid property of ccxml.exit event.
>
> sessionid property is defined as "The identifier of the session."; is not clear which session identifier must be stored in that property, if the current session or the session who has sent the ccxml.kill event.
> In our opinion this sessionid must be the identifier of the session that has sent the ccxml.event and not the CCXML target session, for two reasons:
>
> a) CCXML target session is the one which receives the event itself and doesn't need additional information
> b) If sessionid property of ccxml.kill is different from the identifier of the session which has received the event there is an inconsistency!
>
> Furthermore consequently to this clarification we think that sessionid property of ccxml.kill must be OPTIONAL and not REQUIRED (see attached figure):
>
> ccxml.kill event can be sent in the following ways:
> - by a session to itself (1)
> - by a session to another session which can be a child session or not (2)
> - by an external entity to a session (3)
>
> Case (1)
> Session X send a ccxml.kill event to itself, so the target of<send>  element is X and the sessionid property of ccxml.kill event is X.
>
> Case (2)
> Session X send a ccxml.kill event to session Y so the target of<send>  element is Y and the sessionid property of ccxml.kill event is X.
>
> Case (3)
> An external entity send a ccxml.kill event to session Y: the target of<send>  element is Y but there is no way to assign the sessionid property!
>
> The group approved the following two changes:
>
> So regarding the CCXML spec (http://www.w3.org/TR/ccxml/#flowEventsCCXMLkill) the table of ccxml.kill event can be modified in this way:
>
> - sessionid, not required but optional,
> - Description: The identifier of the session who has sent the event.
>
> =================================
> Paolo,
>
> Could you please clarify the "sessionid in ccxml.kill" issue in more
> detail? Perhaps I'm missing something...
>
> *******
> We are OK with all the remaining resolutions you've posted so far.
> *******
>
> Description of the "sessionid in ccxml.kill" issue:
>
> We understand that the sessionid attribute in ccxml.kill contains the
> session id. The question is who is responsible for inserting the
> sessionid attribute into the ccxml.kill event if the event is sent using
> <send>?
>
> I understand that the platform must insert the attributes listed in
> "9.4.2: Standard Event Attributes" into an event when interpreting
> <send>. The sessionid attribute is not listed here. However, I don't
> think that the platform should insert any other attributes of standard
> events automatically.
>
> Since 6_1.txml doesn't include sessionid in the namelist of the<send>
> it looks like the platform should be responsible for inserting the
> sessionid attribute. Is it correct? It sounds strange to me...
>
> Thanks for clasrification,
> Petr

Received on Friday, 2 July 2010 09:21:11 UTC