I detect a difference in approach between "doing things right" and "doing the least amount of work to get things running". I think both are good ways to make progress, and we should do both.
On Wed, Mar 22, 2006 at 10:49:37AM +0800, Paulex Yang wrote: > There has been a VM/classlib interface definition named as VMI [1] , it > is concise and a *much* smaller interface than *all* native codes > classlib needs. If only the alternative VM implements the VMI and kernel > classes, the sample Java launcher current in SVN should can be used by > it to load classlib native libraries as well as VM libraries, so that > not only java codes but all the classlib specific native codes can be > used on alternate VM without modification, this make sense to me as > *portable*. Yeah it does! (to me) To make my comment above a little more concrete, From a very practical point of view, search-replacing all "native" keywords with '/* native */' is a much lazier/quicker way to get some stuff running (I mean, making JCHEVM implement this VMI sounds like lots of work, and it already implements another abstraction layer (the GNU classpath one), so the best way to do it is probably not so straight-forward), which is not without its merits either. Is there a good reason not to just go and try both ways? Man, who is this guy from the peanut gallery who keeps popping his head around the corner? O wait, its me. Carry on people, carry on... - LSD
