Wow, what a useful thread this is, thanks for all the contributions in unraveling Isis (at least for me)!
I would go for option a. I even like to take it a step further and move out every component in core (released or not) into it's respective folder. For the JDO objectstore the groupId would org.apache.isis.objectstore and the artifactId isis- objectstore-jdo It's still hard for met to get why things are under core or not and what's used in every deployment (imho equals core) and what is used in development. Given that thought my ideal top-level folder layout would be something like this: core/ applib bytecode-cglib bytecode-javassist core runtime development/ tck unittestsupport integtestsupport webserver objectstore/ inmemory jdo nosql sql xml profilestore/ inmemory sql xml progmodel/ java groovy wrapper security/ ldap sql noop file viewer/ bdd junit html dnd restfulobjects scimpi wicket archetype/ dnd-xml scimpi-nosql wicket-restful-jdo retired/ core/ monitoring Still not sure if we need the retired folder though, releasing is all about picking the right combination of modules right? Cheers, Jeroen On Sat, Dec 1, 2012 at 4:36 PM, Dan Haywood <[email protected]>wrote: > On 1 December 2012 14:23, Minto van der Sluis <[email protected]> wrote: > > > >Hmm, I get confused by the artifactIds. I see both for formats > > (isis-viewer-bdd and isis-wicket-viewer). > > > ah, that was a typo. the intention for artifactIds was: isis-xxx-viewer, > isis-yyy-objectstore etc. > > > > I also see artifactIds with 4 > > sections, for instance: isis-jdo-objectstore-metamodel. This confuses > > me. Is it an objectstore or a metamodel? > > > ok, well... some modules (in fact, most) have more than one component. And > I don't think we should insist that all modules have only one component. > > In this particular example, isis-jdo-objectstore-metamodel would be the > additional facet factories that are added to the metamodel that interpret > JDO-specific annotations. That is to distinguish from, say, > isis-jdo-objectstore-datanucleus, which is the stuff that calls the > DataNucleus' specific stuff. > > But there is similar layering in isis-scimpi-viewer, > isis-restfulobjects-viewer, isis-wicket-viewer, isis-sql-objectstore. > > > (isis-metamodel-jdo) Also is > > there really a difference between model and metamodel? > > (isis-jdo-objectstore-metamodel vs isis-wicket-viewer-model). > > > > Yes... there is (though even if there weren't, I'm not sure it matters too > much ... I'd rather give the authors of individual components/modules some > latitude in how they name the individual submodules. For example, scimpi > has isis-scimpi-dispatcher and isis-scimpi-servlet. Anyone who wants to > get involved in enhancing scimpi would grok these particular names and why > they were chosen easily enough) > > To answer your question, though... jdo's metamodel submodule is as > described above, its contributions to the Isis metamodel, whereas wicket's > model is Isis' implementation of Wicket's IModel interface. > > > > > By the way my preference here is isis-objectstore-jdo. For the same > > reason as for the proposed location (easy component grouping when > > viewing directory). In my own projects I usually have my directory > > location match my artifactIds. > > > > ok, so we now we have a couple of distinct and different preferences in > the community. > > Anyone else have an explicit preferences: > a) for the artifactId: isis-jdo-objectstore vs isis-objectstore-jdo > b) for whether artifactId = directory name > > > > > > Also, isis-scimpi-viewer needs some more attention. > > > > the artifactIds weren't correct ... now fixed > > > Thanks > > Dan > > > > > > > > > Please check out the updated version of that wiki page [1] and let me > > know > > > your thoughts. It's important that we get this right (I don't want to > > have > > > to do it all over in 3 months time!!!) > > > > > > Dan > > > > > > [1] > > > > > > https://cwiki.apache.org/confluence/display/ISIS/Make+releases+easier+and+more+frequent > > > > > > > > > > >
