Hi Dan...
--On Friday, May 19, 2000 3:09 PM -0400 Dan OConnor <[EMAIL PROTECTED]>
wrote:
> 1. You forgot to implement a matching ejbPostCreate but we can
> work around this for you, because we're forgiving. WRONG. We
> should instead print out a _helpful_ error message and barf. We
> should not deploy a bean with this type of error, even if we can.
Completely agreed. My primary objection to the nullPointerException was not
that the server barfed if I screwed up, but that it didn't tell me that I
screwed up and how. The error message should be as helpfull as possible,
but I don't need or want the server to write my code for me.
> 2. You forgot to specify the JBoss-specific stateless FIFO pool
> thingy. In fact, you didn't configure anything. We will do everything
> in our power to make your bean(s) work anyway. We were talking
> about how to enable this when the whole "hide the error" thread
> started. Not configuring something is not a real error. Our ability
> to run a bean without configuration does not mask any problems
> inherent to the bean. Our ability to strip away layers of indirection
> when they are not needed does not change the portability of that
> bean for a server that requires you to deal with these layers of
> indirection.
OK, I'll go for this, but I think that when this happens the server should
log that there was no configuration found and it should log what settings
it used. I think this is very important. In fact, can we have a properties
file to set defaults for this event?
> 3. You misconfigured the JBoss-specific stateless FIFO pool
> thingy. Now, this is a more interesting question than not
> configuring it at all. We obviously shouldn't exit with a
> NullPointerException. Should we refuse to deploy the bean until
> you get it right? Should we make a "best guess" and start things
> up, with an error message in the log? Keep in mind that this is
> not a "portable" error. You can't take this type of error and run it on
> WebLogic. :-) Every container will have their own custom
> configuration, separate from the portable core of the EJB:
> deployment descriptor, home interface, remote interface, and
> various implementation classes.
>
> I'm not sure what the best solution is for case 3. I'm pretty sure
> about 1 (barf) and 2 (run), although I'm always anxious for opposing
> viewpoints.
I think this should be an error that causes an abort of deployment, with
very descriptive error messages in the log. This comes close to falling
under the heading of writing my code for me. I would not want to deploy a
bean one way and have it run another, even if (maybe especially if) I
deployed it wrong. I'm happy to fix my mistakes if I can figure out what
they are.
Thanks Dan!
Jim
>
> -Dan
>
> On 19 May 00, at 11:43, marc fleury wrote:
>
>> Jim,
>>
>> <no offense meant>
>>
>> You were the "prototypical" newby user about 3 weeks ago ;-)
>>
>> Don't tell me you want 50 messages telling you, "[ERROR] you forgot to
>> configure the stateless FIFO pool", since we had problems even in the
>> standard ejb-jar.xml and classpath settings (again no offense, i did
>> the same 2 weeks ago).
>>
>> Getting any software to run in EJB is daunting enough that we should
>> make the path "easy" (a la "naive" jBoss 1.0) instead of asking you to
>> answer all these questions. You don't care and you don't want to care
>> about the questions.
>>
>> The features are for the power user (you will certainly become given
>> your fearlessness) and the startups that need a custom adaptable
>> container.
>>
>> </no offense meant>
>>
>> regards
>>
>> marc
>>
>>
>> Jim Archer wrote:
>> >
>> > I agree with this assessment. I hade the same thoughts when I first saw
>> > these messages, but just didn't have time to comment.
>> >
>> > Thanks, Mats.
>> >
>> > --On Friday, May 19, 2000 5:35 PM +0200 Mats Lofkvist <[EMAIL PROTECTED]>
>> > wrote:
>> >
>> > > Regarding 'smart fallbacks' for missing values, I would prefer
>> > > a clear error message stating exactly what information is missing.
>> > >
>> > > This makes the error being detected (and hopefully corrected)
>> > > immediately instead of popping up from nowhere e.g. when moving
>> > > the deployment descriptors to another container.
>> > >
>> > > All 'features' that hide my errors are bad, imho.
>> > >
>> > >
>> > > /mats
>> > >
>> > >
>> > >
>> > > --
>> > > --------------------------------------------------------------
>> > > To subscribe: [EMAIL PROTECTED]
>> > > To unsubscribe: [EMAIL PROTECTED]
>> > > Problems?: [EMAIL PROTECTED]
>> >
>> > --
>> > --------------------------------------------------------------
>> > To subscribe: [EMAIL PROTECTED]
>> > To unsubscribe: [EMAIL PROTECTED]
>> > Problems?: [EMAIL PROTECTED]
>>
>> --
>>
>> ----------------------------------------------------------------
>> Visit Telkel at JavaOne(SM), Sun's 2000 Worldwide Java Developer
>> Conference(SM)
>> June 5-9, 2000 - Moscone Center - San Francisco
>>
>>
>> --
>> --------------------------------------------------------------
>> To subscribe: [EMAIL PROTECTED]
>> To unsubscribe: [EMAIL PROTECTED]
>> Problems?: [EMAIL PROTECTED]
>>
>
>
>
>
> --
> --------------------------------------------------------------
> To subscribe: [EMAIL PROTECTED]
> To unsubscribe: [EMAIL PROTECTED]
> Problems?: [EMAIL PROTECTED]
--
--------------------------------------------------------------
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Problems?: [EMAIL PROTECTED]
