I totally agree with kvale (++) -- highest preference to using Text::xSV on this problem. And just in case you'd like to look at an alternative approach (just as an educational exercise), consider this:

$_ = 'Jul 15 15:12:10 a=foo time="2004-07-14 01:20:25 UTC" b=abc@foo.c
+om msg="^one can say anything here except quotes$"';

my ($timestamp,%pairs) = split(/\s+(\w+)=/);

print "Timestamp => $timestamp\n";
print "$_ => $pairs{$_}\n" for (sort keys %pairs);

__OUTPUT__
Timestamp => Jul 15 15:12:10
a => foo
b => abc@foo.com
msg => "^one can say anything here except quotes$"
time => "2004-07-14 01:20:25 UTC"
[download]

The magic is in using split with a capturing regex, which will divide up the string into its component values, while keeping (and passing along) the attribute titles as well.