Xref: utzoo comp.lang.scheme:1598 comp.lang.functional:407
Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!cs.utexas.edu!rice!news
From: wright@gefion.rice.edu (Andrew Wright)
Newsgroups: comp.lang.scheme,comp.lang.functional
Subject: Re: call/cc and referential transparency
Message-ID: <1990Aug29.145523.286@rice.edu>
Date: 29 Aug 90 14:55:23 GMT
References: <JINX.90Aug22120120@chamarti.ai.mit.edu> <583@array.UUCP> <1990Aug24.212055.8368@nixtdc.uucp>
Sender: news@rice.edu (News)
Organization: Rice University, Houston
Lines: 16

In article <1990Aug24.212055.8368@nixtdc.uucp> doug@nixtdc.UUCP (Doug Moen) writes:
>...
>The conclusion is that call/cc is non-functional no matter how you use it.
>A purely functional language cannot have call/cc or a traditional
>exception handling mechanism.

I agree completely, although I think a simpler argument can be made.
call-with-current-continuation grabs the current continuation, an object
that depends on context.  so obviously referential transparency cannot hold.

Exception handling mechanisms may be viewed as a combination of call/cc
and temporary assignment (or fluid-let), so by extension exception handling
is not "functional" either.

Andrew K. Wright        Computer Science, Rice University
wright@rice.edu         Houston Texas  77251-1892