@Araq: Agree with all your points, I now see there should be few pain points to 
adopting the new memory management model and, as one who has tried to use the 
current GC's system cross thread, very very much to be gained; perhaps the 
primary pain point is that sink move semantics have changed so that the caller 
is in charge of destruction father than the callee - as noted a necessary 
change for consistency and workability in some edge cases but that is more in 
the nature of a design error in version 1:

> However, I'm not thinking too hard about this problem because it's further 
> away...

Assuming by "further away", you mean the implementation of the new memory 
management system, do you have any estimate of how long before there is some 
trial development implementation we can try?

As important as it is, it would really be great to have it included in version 
1.0

Reply via email to