JBossDeveloper

zhfeng In case of the use ${jboss.node.name} I think the presented risks is not the same as for using autogenerated UUID.

> 1. we have no way to know if the "jboss.node.name" or the other UUID is unique

This is not true as using ${jboss.node.name}" has the same properties as when we use value "1". Use of it has rather benefits. In case of "1" we know it's not unique. In case of "jboss.node.name" we help user to differentiate two eap nodes at least by the place it's run. User has freedom to set it up on his own value in the transaction subsystem which will be unique only for transactions. Using the jboss.node.name has the benefit it's known and referred in documentation for user being able to run a cluster or correctly use ejb remoting. We just apply this practice and user knowledge for the transactional subsystem.

> 2. we can not warn the user as this is the default value.

This is what I address in my proposal here: wildfly/TransactionSubsystemRootResourceDefinition.java at master · wildfly/wildfly · GitHub

> Also it can not set the default value of node-identifier as the expression ${jboss.node.name} directly. The possible way could be introduce an attribute with enableAutoGenerateNodeId.

I don't assume to use any auto generated node id. I want to use the ${jboss.node.name} expression to enhance user experience with use of the globaly known WFLY settings in usecases where multiple EAP nodes (clustering, remoting) are started.

So you just only suggest to replace the default value of the NODE_IDENTIFIER with the "${jboss.node.name}". This needs to be introduced in the new version of the transaction subsystem xsd and also need to add the transformer.

A user would have to change from jboss.node.name if they have more than one instance on a host, is that likely deployment topology?

If we were to warn the user when the node identifier was set to the host name perhaps some users would find it confusing to warn on that because they _know_ there is a single instance and it is intentional so the warning would maybe then need some way to be suppressed.

I think we should leave it as it is is as it is so critical to make sure a human makes sure this is unique.

Not at all. Check my proposal: jboss.node.name as default for node-identifier in xml · ochaloup/wildfly@832a0f1 · GitHub  (I'm sorry I put here a wrong link last time)

Ok, as you wish.

Still my argument is that users are used to use jboss.node.name which we can benefit with it and warning would still announce the node identifier should be unique (if it's just left on the value of the hostname).

If you encode the UUID as base64 instead of its native format then 128 bits fits inside of 22 characters.