Interoperability Minutes 2009-04-13

From MemberWiki

Jump to: navigation, search

URL: http://www.openajax.org/member/wiki/Interoperability_Minutes_2009-04-13

Contents

Attendees

  • Jon Ferraiolo, IBM
  • Javier Pedemonte, IBM
  • Howard Weingram, TIBCO

Original Agenda 

Hub.isConnected
Hub.unsubscribe
Hub.getSubscriberData
Hub.getScope
...
I suggest we keep HubClient.disconnect() as is but change the API to say 
that calling any Hub interface functions other than Hub.isConnected will 
result in a Disconnected error if the Hub instance isn't currently 
connected.

That's fine

First, there is a conflict between "Hub.getScope( subID )" and 
"ManagedHub.getScope()".  The former gets the scope associated with a 
given subscription.  The latter gets the scope passed in to the ManagedHub 
constructor.

Second, the documentation for "onSecurityAlert" states that it runs within 
the scope defined by "Hub.getScope()".  This suggests that 
"Hub.getScope()" should function more along the lines of 
"ManagedHub.getScope()".

I suggest the following APIs:
        ManagedHub.getScope()
                - returns "scope" passed in to ManagedHub constructor
        HubClient.getScope()
                - returns "scope" passed in to HubClient constructor
        Hub.getSubscriberScope()
                - returns "scope" passed in to Hub.subscribe()
                - replaces existing Hub.getScope()

Minutes

Issue 1: Hub interface issues with disconnected state

RESOLUTION: As discussed in email, throw disconnected errors

(Javier has already changed the code and the spec. )

Issue 2: move isConnected() from Hub to HubClient interface?

Javier: Resolved in email

RESOLUTION: Leave isConnected() on Hub interface

RESOLUTION: Throw errors immediately for errors that don't require async processing

Javier: There might be a couple of detailed differences remaining where the code and spec aren't completely in sync

ACTION: Javier and Howard to make minor changes to exceptions as needed

Issue 3: getScope() inconsistencies

(discussed in email)

RESOLUTION: Solve problem by having both Hub.getScope() and Hub.getSubscriberScope()

Issue 4: onComplete mandatory / HubClient.getID()

Howard: I'm OK with it being optional. I did some coding this weekend and saw why it is OK to be optional.

(discussion about how this means it is not compelling to switch order of parameters, and how leaving parameters as they are now means first parameters match for OpenAjax.hub.subscribe() and HubClient.subscribe().)

RESOLUTION: Leave onComplete as optional, don't change order of parameters on subscribe()

Issue 5: onComplete callback in Hub and HubClient interfaces

(conclusion that this overlapped with other issues discussed on this phone call, and therefore has been resolved)

Test suite status

(Javier has code for test code for most of the interfaces. Jon has extended Javier's iframe test harness to allow testPubSub.html to be run in a loop against ManagedHub, IframeHubClient, and InlineHubClient, with careful attention to orchestrating timing to deal with async message passing complexities. Jon reports problem where cannot run iframe tests in a loop more than twice, which is mysterious since the DOH engine treats each test as a separate independent web page. Maybe aggressive browser caching is causing problems. Jon to package up the problem and send to Javier to review.)

Retrieved from "http://www.openajax.org/member/wiki/Interoperability_Minutes_2009-04-13"
Personal tools