From: Gavin_King/Cirrus%<CI...@ci...> - 2002-07-17 15:07:28
|
>I think you'd have to get the file's timestamp and then poll it at regular >intervals to see if it changed. I wanted to do that with the Hibernate >MBean I created but didn't have the time. Hmmmm....not totally convinced on this. Don't like the whole concept of polling the hard drive actually. I *can* be convined though.... >> That is the recommended way to deploy mappings, as far as I'm concerned. So we would definately allow this. << >Hmm. I've been assuming the use of one big mapping file. I supposed if you >put each class mapping in a seperate file, that would take care of it. It >would also make the part about checking for the mapping file changing more >difficult. I really like the idea of having little mapping files living in the same directory/jar as the class they map. I should more strongly recommend this in the documentation. Monolithic mapping files are much less elegant. >>As far as I can tell the only JBoss-specific thing in your MBean is the startService() / stopService() methods from the inherited JBossServiceMBean interface. (Since JBoss has a compatible license, we should be able to just distribute that interface along with Hibernate.) Then, as far as any other JMX client is concerned startService() / stopService() are just any old management operations that could be invoked by the user (rather than by the container in the case of JBoss).<< >How would the service get started in non-JBoss containers? I hadn't quite made up my mind on that yet. From a JMX client possibly. Though that would require user-interaction.... P.S. I realised my mistake above. Your *implementation* class actually inherits some JBoss-specific stuff. Ima have to track down that code to see what it does. Anyway, since JBoss is the only appserver that really supports JMX, I'm thinking mainly about non-application-server usage. It may turn out we need one MBean for JBoss and one for other JMX applications. Okay, let me get back to learning JMX, after I fix this damn bug in 1.0.1 :( |