11.3 resource allocation

[Mike Prince]

On Wed, 7 Dec 1994, Francois-Rene Rideau wrote:
> Let's have high-level abstractions to represent communication; not low-level
> implementation. Then, our standard libraries can map those abstractions
> to stacks or anything actually available with best performance.

I agree it's nice to have high-level abstractions, but we also need to 
specify the exact message format between systems, so what comes from mine 
onto that floppy disk, will boot on your Cray.


> In continuation-passing style, you have no stack, but call/return is just
> a jump instruction; the destination should handle whatever is needed to
> continue computation (which can be using a stack or not).
Which brings us back to "How do we get home?".  We are basically stuck 
with a stack, whichever form it's in.  Our problem is how do we represent 
it in such a way that the data is good even after migration has caused 
all our return addresses to become invalid.

Mike