[bitc-dev] Region-based Memory Management?

[Jonathan S. Shapiro]

On Sat, 2005-01-29 at 20:54 +0100, Jørgen Hermanrud Fjeld wrote:
> I wondered about the relevance of region based memory managment [1] to
> BitC, as in the work on memory regions in Cyclone [2]. 
> Cyclone is an extension of C that provides memory region
> annotations and memory region inference such that one can avoid 
> dangling pointers.  
> If you must provide pointers, how would you then prevent dangling
> pointers, if not by some mechanism such as memory regions? 
> 
> [1] http://www.research.att.com/projects/cyclone/papers/cyclone-regions.pdf 
> [2] http://www.research.att.com/projects/cyclone/

I am familiar with Cyclone, and have actually tried to write some non-
trivial code in it. Speaking as a developer, my experience with region
based memory management was *horrible*.

The best answer to your question is (a) verification, or (b) garbage
collection.