At 10:36 PM 8/8/00 -0400, Ken Fox wrote:
>Dan Sugalski wrote:
> > At 09:25 AM 8/6/00 -0400, Ken Fox wrote:
> > > We're definitely going to have some sort of traversing collector. It
> > > might only augment the ref count collector.
> >
> > I'd rather not have multiple GC schemes going at once. (Though we may need
> > some sort of "refered to by" field to track variables that have had
> > references to them taken, so we know when the last referent goes out of
> scope)
>
>I can say right now that mark-sweep won't work by itself. Neither will
>ref count. Combine them together and it might work well enough. (Perl also
>needs some way of dealing with young data which suggests that a special
>stack allocator with write barriers is needed.) If we abandon ref count, then
>we also need to make sure the new collector is ephemeral. People will *not*
>be happy if their scripts visibly pause to GC. (Billions of invisible
>pauses like we have today go undetected... ;)
Y'know, I think my head's going to hurt a lot once Quantum gets that GC
book I ordered...
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
