[Dev] Re: refcounted

[Andi Vajda]

> That said, I think his choice of the default will not make Chandler code more 
> portable (it's using refcounted) or change people's preference for using

That's correct, it's not my choice which determines how portable Chandler is.

The whole argument - somewhat bogus - of porting to Java, has been shadowing 
the real point I was trying to make about portability in general.

Pedronis, on IRC, just added this to the debate. If a faster python VM were to 
be made, would Chandler not try to take advantage of it ?

   Pedronis: hi, I see you are debating about refcounting
   Andi    : we're not debating refcounting anymore, we're debating planning
             for nimbleness but did you want to add something to the
             refcounting debate, pedronis ?
   pedronis: one thing to consider longterm is that future python versions
             with better performance might not be ref counted at all
   Andi    : we're all in agreement that, for Chandler, in python, the
             refcounting feature is a gain, I don't think there is any
             disagreement about that. I think this feature is very cool, I'm
             proud of my implementation even.
   heikki  : pedronis: have there been any plans to change python like that
             - it seems a really fundamental change to do?
   Andi    : so, pedronis, is the sys.getrefcount() method going to disappear ?
   pedronis: no CPython is going to stay that way
   pedronis: but over the next 2-4 year other python versions have a
             non-zero chance to appear
   Andi    : so porting Chandler to faster python VMs may be problematic with a
             hard dependence on refcounting, you're saying ?
   pedronis: with far better performance and likely not ref counted
   Andi    : would these later VMs support weak references ?
   pedronis: yes, I think so
   Andi    : I was thinking that another way to implement the same affordance
             as 'refcounted' is by way of weak references but I haven't
             investigated that too much.
   pedronis: anyway the faster Python is a possibility, nothing sure, but there
             are various effort starting or ongoing in that direction