On Mon, Jun 16, 2008 at 6:36 AM, Rickard Öberg <[EMAIL PROTECTED]> wrote:
> Moving all that into SPI would expose a LOT of stuff. One thing we could > do is to continue adding *Descriptor classes in SPI, and let the > ModelVisitor work with that instead of the raw *Model classes. A Visitor > wanting to get to the runtime can cast the *Descriptor classes anyway I > suppose. How about that? Well, I see a few goals here, in priority order; 1. No code at all depend on Runtime. (Soon this will be done for bootstrap as well.) 2. As little actual code as possible in API and SPI. Prefer interfaces over classes. 3. No complex instantiation patterns slowing down the 'critical paths'. So, IMHO, if the RDFizer and/or its ModelVisitors needs to cast to the runtime classes, then the pattern is flawed, and we should fix that. One part of this 'puzzle' is of course to define what are the formal structures that goes into the RDF description, as these should be maintained. Cheers Niclas _______________________________________________ qi4j-dev mailing list [email protected] http://lists.ops4j.org/mailman/listinfo/qi4j-dev
