[racket] Callback on shutdown

From: Matthew Flatt (mflatt at cs.utah.edu)
Date: Tue May 6 17:12:53 EDT 2014

At Mon, 5 May 2014 11:53:09 -0400, Greg Hendershott wrote:
> Would it make sense for will executors [^1] to have a definition of
> "unreachability" that includes GC, as now, but also immanent program
> termination (and they run automatically on termination)?

I don't think that would work well.

Suppose that you have wills on objects A and B, where B is reachable
from A's will (i.e., the will for A assumes that it can use B). Then,
it would be important to execute A's will before B's will. Collecting
that kind of dependency information is difficult and/or expansive.

Also, will executors do not form a hierarchy in the same way as
custodians. A will executor can itself become unreachable, in which
case none its remaining wills are executed.


Posted on the users mailing list.