Mailing-List: contact continuum-dev-help@maven.apache.org; run by ezmlm
Precedence: bulk
Reply-To: continuum-dev@maven.apache.org
Received-SPF: neutral (athena.apache.org: local policy)
Mime-Version: 1.0 (Apple Message framework v752.3)
In-Reply-To: <a43524d00707311901h6844d8e9sad238c16d3b9f95b@mail.gmail.com>
References: <4F93F9D0-BE70-43A7-83FE-24D96322ACF3@apache.org>
 <a43524d00707311901h6844d8e9sad238c16d3b9f95b@mail.gmail.com>
Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed
Message-Id: <1888241C-2921-437A-8C65-40567E8B992A@apache.org>
Content-Transfer-Encoding: 7bit
From: Brett Porter <brett@apache.org>
Subject: Re: Solving the notification problem
Date: Wed, 1 Aug 2007 12:10:31 +1000
To: continuum-dev@maven.apache.org


On 01/08/2007, at 12:01 PM, Jesse McConnell wrote:

> well, 1.1 is in beta now so ideally no more features, but maybe we  
> just call
> this a bug fix for now..

yeah, I'm looking more for a small improvement than a revolution.  
Avoiding schema changes and such. I really should have said something  
before beta-1.

>
> I guess the first suggestion might be a batch email option with a  
> summary
> presented at the top of the mail...that would almost have to be on  
> a per
> notifier though, but that is made easier by the group level inherited
> notifiers...That is another db schema change though :)

Yeah, and we don't really have anything that can keep track of when  
to send that - there's no notion of a group schedule.

>
> We could alter the notifications to prepare web reports that have  
> the links
> mailed out, and suppress mails based on time or if a given link has  
> been
> mailed recently.

We already have the web reports, so maybe a configurable time  
threshold is the best (ok, it might require an addition to the  
notifier schema, but it's innocuous.... :)

>
> what conditions do we find the highest lvl of easily ignorable  
> mails go out?
>
> I can think of things like SCM outage and things like that...out of  
> memory
> errors...

Errors are definitely the biggest nuisance, and probably chain  
reaction failures. If we could even take care of the error condition  
stuff it might be enough.

- Brett