Hi Justin,

Thanks for all the sophisticated ideas regarding the next Ruby GC.

I am still waiting for Matz's response regarding whether it is possible to
simply use "rb_gc_un/register_address()" (or a similar concept) for GCed
objects that are referred to from non-GCed objects.  If such a concept
is indeed possible, then one feature that I would like to have in Ruby GC
is a separation of GCed objects from non-GCed objects.  With the existence
of non-GCed objects, then we will have tighter memory control just like
that in ordinary C programming (and this will make Ruby more ahead than
the other scripting languages, where at least in theory, the Ruby speed
can be made arbitrarily close to the C speed).

I always think that the only reason people write Ruby in C is, beside to
interface with external C libraries, to increase execution
performance.  Currently, the Ruby GC is one of the performance bottlenecks
and even worse, it does not scale well.  Therefore I am really glad that
there are people like you who are considering the Ruby GC very seriously.

I am always dreaming of "The speed of C, the simplicity of
Ruby"...  Hopefully it can come true...

Regards,

Bill
=============================================================================
Justin Johnson <justinj / mobiusent.com> wrote:

> Which method of garbage collection would most suit Ruby?

(...deleted...)

> Anybody out there with any views on this?

> Matz, did you have any ideas on the GC for Rite?

> --
> Justin Johnson