Hi,
It seems that OASIS version of the spec provides a direction here. See: http://www.osoa.org/jira/browse/ASSEMBLY-44. The proposal is quoted below: "All of these artifacts within an SCA Domain are defined in SCA contributions in files called META-INF/definitions.xml (relative to the contribution base URI). Although the definitions are specified within a single SCA contribution, the definitions are visible throughout the domain. Because of this, all of the QNames for these definitions MUST be unique within the domain. " I personally think it might be better to provide a bit scoping based on the import/export statements. But it would be a question to the spec people. Thanks, Raymond From: Simon Laws Sent: Monday, January 12, 2009 2:15 AM To: [email protected] Subject: Re: [1.x][2.x][DISCUSS] definitions.xml processing Hi Greg Comments in line... Simon On Fri, Jan 9, 2009 at 10:38 PM, Greg Dritschler <[email protected]> wrote: Simon, Regarding bullet A, this is fine if the set of contributions is static. What happens as contributions are added and removed (particularly contributions that contain definitions)? I think the spec is silent about what should happen. I think the confusion is because the spec doesn't state that definitions.xml will be provided via contributions. I think having definitions.xml inside contributions is a Tuscany thing. The spec talks about a domain wide file named definitions.xml. Hence it makes no comment on what happens as contributions are added and removed. Given that we allow definitions.xml files inside contribution I was looking at what it would take to really consider these definitions artifacts as domain wide. The answer is that we have to change our read and resolution process a bit but it's not a huge change. Another approach would be to change the way Tuscany works and read definitions.xml files independently of contributions or, for example, mandate that a separate system contribution must be provided to add defintions.xml to the domain. As you point out, I haven't addressed what should happen as contributions are added and removed in a running syste. In fact we currently don't address this generally. The Tuscany domain is static in that all contributions must be present before you start nodes. I'd like to see us move toward a more dynamic solution but we aren't there yet. Any change in contributions (or defintions.xml) within a running system would imply some kind of eventing to notify running nodes that their configuration has changed. This is obviously a rather complex proposition given where we are now. Give where we are with 1.x I think the answer is to; - fix our read/resolve process so that definitions.xml are considered to be domain wide - fix our domain manager so that the domain definitions.xml model is passed to each node Given these two changes we can close off TUSCANY-2499 - make a statement that once contributions have been added they can't be removed without restarting the nodes that refer to them. If you are prepared to restart a node then you are presented with the domain configuration as it appears at that time taking into account whatever contributions have been added or removed. For 2.x we could either apply the changes to make definitions that appear in contributions be processed as domain wide defintions or take a wider view and choose a different approach to reading in definitions.xml. Raymond, Are you suggesting that the user should have to explicitly import definitions from other user (non-system) contributions? As far as I know all definitions that have been contributed to the domain are implicitly visible to all contributions. I agree. I don't think the user should have to import artifacts from definitions.xml files. The spec refers to them as domain wide artifacts.
