cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gianugo Rabellino <gian...@apache.org>
Subject Re: [RT] On building on stone
Date Mon, 29 Mar 2004 09:16:41 GMT
Ralph Goers wrote:

> I'm curious why you think OpenMBeans are awkward and unmanageable.  Seems
> pretty easy to me. What am I missing?

Well, there is a principle in subatomic physics (but, hey, I'm a lawyer 
so don't get picky just now :-)) saying that even observation can cause 
perturbation of a system, so that measurements are not to be trusted 
beyond a certain extent since the act of measuring might have impacted 
the results.

I'm afraid that this is the case with "advanced" JMX. I'm still a newbie 
on different implementations, but I think that instrumentation should be 
a *thin* layer on top of business objects (and this seems to be thee 
case for standard mbeans). When you enter dynamic mbeans the situation 
gets nasty, since you have to support a whole mess of object graphs 
(MBeanInfo having references to MBeanConstructorInfo and 
MBeanOperationInfo, which in turn have references to MBeanParameterInfo, 
plus MBeanAttributeInfo and MBeanNotificationInfo: hey, wait, does that 
mean that to manage _one_ freaking object I have to have roughly 10 
support objects? Now come on...). Model and Open MBean keep adding 
complexity since they are dynamic beans plus "something", raising the 
object count to about 15 for every managed object. Now, get real...

Now, I'd really love to be proven wrong, but heck, this is a serious 
mess. What are the chances of having/introducing bugs in the management 
part alone? And what is the introduced overweigth of such management?

Ciao,

-- 
Gianugo Rabellino
Pro-netics s.r.l. -  http://www.pro-netics.com
Orixo, the XML business alliance - http://www.orixo.com
     (Blogging at: http://www.rabellino.it/blog/)

Mime
View raw message