Re: [Swarm-Support] Re: Swarm 2.2 on Intel (and PPC) Mac OSX Tiger

[Marcus G. Daniels]

Bill Northcott wrote:
> Having said this, it seems to me to be a good idea to eradicate them 
> now from the Swarm code base rather than wait until something breaks.
While I'm happy to see Scott's helpful #ifdef workarounds in CVS for 
this, I'm not convinced at all it improves Swarm to code in this way.   
In some cases he had to pass around nearly a dozen parameters.   Nested 
functions facilitate passing functions as arguments, capturing state 
that is intended to be invariant for that iteration.  It implements 
lexical scope for C.  It's both efficient and easy to understand what is 
intended because the iterated action is defined in the immediate 
context, and only the parameter that changes is passed.  Objective C is 
a reasonable place to use these kinds of C extensions since Objective C 
is not ISO 99 C.   As far as worries about an stack exploit go (which 
assume that someone wants to run models from web servers or as root), 
those opportunities are reduced by the compiler turning on and off stack 
executability in narrow windows of time.

To make this kind of change in Swarm as a preference and not just a 
workaround, I'd suggest making structures for the invariant state and 
passing that single extra parameter around.   Another option would be to 
reorganize the code around objects and pass method selectors around, or 
their associated pre-dispatched IMPs.

To disable this feature sounds fishy and a bit fascist to me, like "when 
something is hard or confusing, forbid it". 

On the other hand, having a portable pure C or C++ implementation of 
Swarm interfaces would be great but that's a completely different argument.

Marcus