On 9/12/07, Simon Laws <[EMAIL PROTECTED]> wrote: > > On 9/12/07, Simon Nash <[EMAIL PROTECTED]> wrote: > > > > Comments inline. > > > > Simon > > > > Simon Laws wrote: > > > > > Thanks for the reply Sebastien > > > > > > A few more comments below... > > > > > > On 9/12/07, Jean-Sebastien Delfino <[EMAIL PROTECTED]> wrote: > > > > > >>Simon Laws wrote: > > >> > > >>>There are some reorganized domain/node classes under > > modules/distributed > > >> > > >>and > > >> > > >>>modules/distributed-impl dirs. Here the SCADomain is replaced by a > > >> > > >>Node. A > > >> > > >>>node runs all or part of a domain. A Node has contributions added and > > >>>removed and has components started/stoppped etc. If available (a > domain > > >> > > >>and > > >> > > >>>node name are provided and the domain is running) a Node registers > > with > > >> > > >>a > > >> > > >>>DomainManager service and a ServiceDiscovery service. Here's what's > in > > >> > > >>the > > >> > > >>>new code > > >>> > > >>>Node > > >>>- A node implementation (NodeImpl) > > >>>- A contributions manager for adding/removing contributions > > >>>- A component manager > > >>>- A management SCA application that provides access to these features > > >>>remotely > > >>>- a web page for looking at the node > > >>> > > >>>Domain > > >>>- A ServiceDiscovery service > > >>>- A domain manager service > > >>>- A sample domain application that pulls two together and includes > > >>>- A web page for looking at the domain and provides links to each > nodes > > >> > > >>web > > >> > > >>>page. > > >>> > > >>> > > >> > > >>This looks pretty good to me! So far I've looked at the interfaces and > > >>the implementation, and will try the web pages next :) > > >> > > > > > > > > > I'd like to make a proposal to change ServiceDiscovery a bit, but I'll > > > > > >>do that in a separate email. > > >> > > >> > > >>>You can try this by running the calculator distributed sample. This > > runs > > >> > > >>and > > >> > > >>>exercises some distributed nodes as it always has but uses the new > > >> > > >>classes > > >> > > >>>now. If you run the nodes independently from the command line they > stay > > >>>around long enough for you to point a browser at them. Try > > >>>htpp://localhost:8080/node/index.html (or whatever port the node has > > >> > > >>come > > >> > > >>>up on) and see the components in a node. > > >>> > > >>> > > >>>There is a new sample (sample-domain-webapp). The intention here is > to > > >>>provide a generic domain and a node so you can start the domain and > > node > > >> > > >>and > > >> > > >>>then add contributions. Not complete yet as the "add contribution" > > >> > > >>function > > >> > > >>>needs turning into a remote service but you can use the domain > > >>>implementation along with nodes from the distributed calculator > sample > > >> > > >>which > > >> > > >>>have hard coded contributions. > > >>> > > >>>Here are some todos > > >>> > > >>>I've copied all of the interfaces I need to make this work into > > >>>modules/distributed so there is code/interfaces here that is also > > >> > > >>elsewhere, > > >> > > >>>for > > >>>example, the component manager classes. I would like to move the new > > >> > > >>code to > > >> > > >>>new modules > > >>> > > >>>modules/host-node - for the node related code? > > >>>modules/host-domain - for the domain related code? > > >>> > > >> > > >>How about this? > > >>modules/node > > >>modules/domain > > >> > > >>IMO host-* is for the integration with hosting environments like > Tomcat, > > >>Jetty, an HTTP or RMI server. > > >>And host-embedded should probably not be called host-embedded :) > > > > > > > > > > > > Sounds OK to me - I'll go ahead and split them out. > > > > > > > > >>I have used the interfaces Node and Domain currently should this be > > >>SCANode > > >> > > >>>and SCADomain? > > >>> > > >> > > >>I'm OK with both, not sure what others prefer. > > > > > > > > > > > > I'm ambivalent but we have one positive request for SCANode and > > SCADomain so > > > I'll wait a little longer and probably change to that > > > > > I don't think we need "SCA" in front of these names. After all, > > just about everything we are doing in Tuscany is to do with SCA, > > so if we go down this path we might see "SCA" name prefixes > > cropping up everywhere :-( > > > > Is there a reason why these two names would need "SCA" in front of > > them? Do we have any other "Node" or "Domain" concept in Tuscany > > that could be be confused with these? > > > > > > > >>host-embedded can stay as it provides the runtime and embedded support > > but > > >> > > >>>there are numerous domain implementations that we can remove assuming > > we > > >>>get to the stage where we are comfortable with Node. Ant has already > > >> > > >>ported > > >> > > >>>the hot update web app to use the new domain (currently working > through > > >>>an issue with uris) > > >>> > > >>>I'd like to start using the Node implementation in the samples. I'll > > >> > > >>have a > > >> > > >>>go at converting some and see how it goes. > > >>> > > >> > > >>Great! > > >> > > >>I'd suggest to move the API methods (expected to be used in > application > > >>business logic) like getService(), getServiceReference() and cast() to > a > > >>separate interface in a separate domain-api or node-api module. > > > > > > > > > > > > OK, i'll take a look at that. > > > > > + 1 for this. I think the new module should include the API for > creating > > a domain as well. > > > > Simon > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > I don't think we have more than one node or domain concept. These words > are used elsewhere many times (outside of Tuscany) and it could be useful > to > ensure that people understand that we are talking about the Tuscany > concept > of Node and Domain rather than anyone else's. The code at the moment uses > Node and Domain. I raised the question as I felt there was scope for > confusion. I'm now thinking that the SCA prefix is too loose (and not > applicable in the Node code) so maybe TuscanyNode/TuscanyDomain would fit > the bill?
I still prefer SCADomain out of all the suggestions :) For this particular case i think its good to have the SCA suffix and that it makes it more intuitive what its about. ...ant
