On Tue, May 13, 2008 at 7:33 PM, Rickard Öberg <[EMAIL PROTECTED]> wrote:
> Well, in general I would want to keep SPI as small as possible (it *is* > a PI after all), so not sure there's a point to keep that code there > rather than runtime. But either way works. Our general trend is to get > as much as possible into core.runtime, and only "leak" to SPI or API > when we absolutely have to. It's a matter of encapsulation and > maintainability. Yes, it is important to understand the scope/responsibilities of each; * API - The stuff that regular domain applications need to know. * SPI - The stuff extensions and really clever applications can use, kind of the "reflection API" of Qi4j. * Runtime - Internals * BootStrap - The stuff the Qi4j instance controlling code need to get Qi4j up and running. There are a couple of cases where a bit more sit in API than absolutely necessary, but that is due to the dependency graph. What I would like to see is that we can refactor Bootstrap to not depend on Runtime directly. Since this introduces an "annoying" dependency that irritates me. Cheers Niclas _______________________________________________ qi4j-dev mailing list [email protected] http://lists.ops4j.org/mailman/listinfo/qi4j-dev
