CFP: USENIX Java Virtual Machine Symposium

Sat Oct 14 21:09:00 GMT 2000

>> One thing I would like to see published is a description of the
> "constant-time interface" algorithm, designed by me, and implemented
> by Bryce McKinlay. However, to make a good paper, we need some
> numbers and experience, which we probably don't have time for.
> I think this may actually be be worth an OOPSLA paper. (Aiming
> for next year's OOPSLA gives us more time, plus OOPSLA is a more
> prestigious conference, plus the results are not specific to Java.)
> If someone wants to do the experiments, perhaps we could write a
> joint paper (with Bryce and myself)? I repeat: I think this is
> definitely something that *should* be published.

Isn't the algorithm gcj uses basically selector indexed dispatch 
tables (STI) which is pretty standard for multiple inheritance
dispatch? (Except that you don't spend any time trying to
compress the selector tables?)
How is it different?
	- Godmar