RMI support? (fwd)

[Godmar Back]

> 
> Godmar Back wrote:
> > 
> >  Chris, it would be great if you let the list know what you find out
> > with regards to Ninja's compatibility.
> > 
> > You are bound to find bugs in kaffe's serialization implementation, in part
> > because it is incomplete.  However,  I do not think that the problem is lack
> > of knowledge of the native methods that Sun uses to implement their pickling
> > and the like.  Some known problems that are mentioned in the FAQ include
> > that kaffe's object serialization does not follow spec; for instance, it
> > does not compute serialversion uid according to the algorithm described by
> > Sun.
> > 
> > As an aside, I should point out that RMI worked in the last pre-1.0b1
> > snapshot --- when kaffe still used Sun's classes.zip and implemented the
> > native methods by guessing --- so ironically we got those native methods
> > upon which Sun's implementation relies right.  The current problems lie in
> > the Java part, which is no longer based on any code from Sun.
> I should be able to answer this for you early in the new year. Out of
> curiosity, is there any reason why we aren't using Sun's method for
> computing serialversion uid?

It's not yet implemented.  That's the only reason.
We must indeed use Sun's method for interoperability.  The algorithm is
part of the "spec"; Sun's implementation is not and must be recreated.

	- Godmar