Re: exception systems

["Guillermo J. Rozas" <gjr@martigny.ai.mit.edu>]

|   Date: Fri, 19 Apr 1996 14:49:46 -0400
|   From: "ozan s. yigit" <oz@nexus.yorku.ca>
|   Cc: rrrs-authors@mc.lcs.mit.edu
|
|   quick question unrelated to the general discussion:
|
|   i am not sure if i understand what the problem is with SML/NJ call/cc
|   implementation. are there more problems in implementing it for SML than
|   scheme or does SML require a different strategy with different tradeoffs?
|   [if i am missing the last word on this topic in scheme, please refresh my
|   memory. the last thing i read is dbyvig's one shot" paper, sorry.]

There is no problem with their implementation.  They use
heap-allocated continuation frames instead of a traditional stack.
This is controversial at least.

The one uncontroversial (I think) advantage is that it makes cwcc
relatively cheap.  The question is how expensive this decision makes
everything else.

The many incremental stack-copying solutions that I know of make it
acceptable in performance, and do not penalize anything else (with a
few minor exceptions relating to state byt SML/NJ does not take
advantage of these to my knowledge).