local *UNIVERSAL::method = sub {} - odd behaviour

The notion of 

local *UNIVERSAL::method = sub {};

struck me while I was reading Damian's book over the weekend.
(As a way of providing a fallback 'default:' in an OO 'case'.)
 
I am not surprised it 'works' but I am slightly suprised it 'lingers'
the way it does:

#!perl -x
my $obj = bless {},'Foo';

sub harry
{
 local *UNIVERSAL::method = sub { print 'harry ',shift(@_),' ',join(',',caller),"\n" };
 Thing->method;
 $obj->method;
}                            

sub fred
{
 local *UNIVERSAL::method = sub { print 'fred ',shift(@_),' ',join(',',caller),"\n" };
 Thing->method;
 $obj->method;
}                            

harry();               
Thing->method;  # no error !
$obj->method;
fred();         # does get new method!          
Thing->method;
harry();
Thing->method;
__END__

Yields:

nick@bactrian:/home/perl > perl locuniv
harry Thing main,locuniv,7
harry Foo=HASH(0x80de9dc) main,locuniv,8
harry Thing main,locuniv,19
harry Foo=HASH(0x80de9dc) main,locuniv,20
fred Thing main,locuniv,14
fred Foo=HASH(0x80de9dc) main,locuniv,15
fred Thing main,locuniv,22
harry Thing main,locuniv,7
harry Foo=HASH(0x80de9dc) main,locuniv,8
harry Thing main,locuniv,24
nick@bactrian:/home/perl >    

You don't get any errors outside the scope of the local - it re-uses the 
last one - presumably because CV has been cached in Foo::method 
and Thing::method. But if that is the case why does the re-definition 
in fred's local "work" and not call the cached CV? (glad it does!).

Could/should the scope exit be tweaked to make it behave as I expect ?

-- 
Nick Ing-Simmons

• local *UNIVERSAL::method = sub {} - odd behaviour by Nick Ing-Simmons
• local *UNIVERSAL::method = sub {} - odd behaviour by M.J.T. Guy
• Re: local *UNIVERSAL::method = sub {} - odd behaviour by Nick Ing-Simmons
• Re: local *UNIVERSAL::method = sub {} - odd behaviour by Gurusamy Sarathy