FW: [Openais] Certain Things Regarding Event Service API

[Steven Dake]

On Tue, 2005-07-26 at 07:27 -0700, Mark Haverkamp wrote:
> On Tue, 2005-07-26 at 10:07 +0530, Arunachalam wrote:
> > Sir,
> > 
> > 	And about the 2nd question of 'LOST EVENT' event ?
> 
> I disagree that a lost event message should be sent to the entire
> cluster.  The situations causing, and response to, a lost event are
> described on page 23 of the specification.  The way that I understand
> this section is that it applies to individual subscribers.  If an event
> cannot be delivered to an individual subscriber for some reason, the
> LOST EVENT event is sent to that subscriber only. 
> 
> Mark.
> 

Mark
I agree with your assessment of the lost event definition on page 23 of
the evtB spec.  Lost events are sent per subscriber, not per channel.

regards
-steve


> 
> > 
> > Regards
> > A P Sarathy
> > 
> > 
> > 
> > -----Original Message-----
> > From: Steven Dake [mailto:sdake at mvista.com] 
> > Sent: Friday, July 22, 2005 10:04 PM
> > To: arunachalamp at huawei.com
> > Cc: openais at lists.osdl.org
> > Subject: Re: [Openais] Certain Things Regarding Event Service API
> > 
> > On Fri, 2005-07-22 at 16:34 +0530, Arunachalam wrote:
> > >      1. In the Publish API (library part)SA_AIS_ERR_TOO_BIG should be
> > >         returned if the event data size is greater than
> > >         SA_EVT_DATA_MAX_LEN (With reference to SAF Spec B 01
> > >         Page.No:15, Line.Nos: 10-13). Presently the error code
> > >         returned is SA_AIS_ERR_INVALID_PARAM.
> > > 
> > >  
> > 
> > I agree I'll open a bug.
> > 
> > > 
> > > 2. In the implementation of SAF B 01 Spec of  OpenAis 'LOST EVENT'
> > > Event is not delivered to all the subscribers for the particular
> > > channel. The message is sent only to the subscriber pass of the loop
> > > when the lost event is generated.
> > > 
> > I think your right here.  What we need is when a lost even occurs, the
> > event executive should send a "lost event" message via executive message
> > handler.
> > 
> > I leave it up to mark to solve this one though :)
> > 
> > regards
> > -steve
> > 
> > 
>