While agreeing with everything everyone has said above, I thought I might as well :

use warnings;
use strict;

my $string = q(<a style='postion: top;   font:roman' href=hi.html href
+='bold' src=" image" />);

my @tokens = $string =~ /<?\s*(\w+=([\'\"]).*?\2|[^\s>]+)/g;

# Get every other token
my $i = 0;
@tokens = grep {++$i % 2} @tokens;

local $" = "\n";
print "@tokens\n";
[download]

This gives:

a
style='postion: top;   font:roman'
href=hi.html
href='bold'
src=" image"
/
[download]

This will only work with this data format, i.e. with the token=quoted data layout. The RE uses a back-reference for the second quote because you might have quotes-in-quotes. This complicates the result array, hence the grep.

You really are better off using a CPAN module for the general case.