$'name

[mkmcconn]

If the $' in $'var is not at all related to $', how does Perl keep them straight?
ex:

#!/usr/bin/perl -wl
use strict;

$main::var       =  "a rose by ";         #$main::var
$main'var        =  $'var."another name"; #$main::var again

use vars q($var);
$var             =~ s/(\b\w)/uc($1)/ge;  #$main::var 
$var             =~ /N/;
print $'var;                              #print $main::var

print 'G'.$'." With $'var";               #games with $'

use English;
print "s$' and s$POSTMATCH";

mkmcconn

[tye]

See line 5811 of toke.c (for Perl5.6.0):

else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {

So, in effect, /\$'[a-zA-Z_]/ is treated differently than /\$'[^a-zA-Z_]/ [note the caret (^)].
        - tye (but my friends call me "Tye")

[tadman]

I would suppose that the lexer in the perl parser is intelligent enough to recognize the pattern for variables of that sort. It could loosely be described as:      \$'([a-zA-Z_][a-zA-Z0-9_]*)? Which could define a valid variable name.

Otherwise, how would you distinguish between $x and $xy? Or, $_ and $_x for that matter.