From: Charles Oliver Nutter <charles.nutter@...>
Date: 2007-07-18T06:37:43+09:00
Subject: Re: Proposal: runtime-modifying Kernel methods should be keywords

Brent Roman wrote:
> Just a follow up to on the idea of disallowing the
> aliasing of :eval and similar methods...
> 
> Rather than making this a static switch specified
> apriori, could the VM / compiler agressively optimize
> until the "internal" methods (:eval, etc.) were
> aliased/redefined.  When an alias or redefinition
> was detected, it would output a warning to the
> user that performance will suffer and disable
> the problematic optimizations.
> 
> Or, would this also require regenerating code for all
> the previously compiled methods that invoked
> :eval and their ilk?

It potentially would require that, and there's also the complication 
that we may not be able to detect if it's been aliased, if say there's a 
call to an aliased "alias_method" for example :) The rabbit hole deepens.

I think the aliasing restriction isn't a bad one..there's already a 
similar restriction on aliasing some global variables like $1 since 
they're handled earlier in execution.

- Charlie