|
From: Bordet, S. <Sim...@co...> - 2002-01-28 09:18:38
|
Hi Tom, > I have been working with the JMX Reference Implementation to=20 > build some > additional features (similar to the JDMK). I have everything=20 > working using > the JMX RI and found your implementation. =20 Be aware that coding against the RI is very different from coding = against the JMX spec. Many threads have been discussed here that start = with "cannot run this in OpenJMX, it runs in the RI, why ?" and that = ends up with "a bug in the RI". > I found that your version of the > NotificationBroadcaster interface has 3 method signatures for > removeNotificationListener() and the JMX RI implementation=20 > has 1. After > looking at the JMX spec., it says that: >=20 > "This method takes a reference to a NotificationListener=20 > object, as well as > a hand-back object." >=20 > The JMX RI implementation isn't even correct, in it only takes the > NotificationListener as a parameter. I know the specs. are vague, but > wondering why you added the extra ones? Because they are very much needed :) Furthermore, the spec should upgrade and fix this mess: the only way I = see not to break the already written code is to add the spec compliant = method and/or the third one, that was missed by the spec writers, but is = indeed needed. For further info, you can look in the archives of this mailing list, or = in the ones of jmx...@ja..., where this issue has been = discusses in detail. > P.S. Also wondering if there are any future plans to add=20 > some of the extra > features as in the JDMK to OpenJMX in the future, or will you=20 > wait till > solid specs. are release on these types of features (i.e.=20 > Cascading Agents, > auto-detection, etc.)? For what pertain JSR 160, I guess we will wait the spec, even if I'm on = the expert group for that JSR. For other feautures such as SNMP support and other stuff, we are wide = open to contributions. Regards Simon |